Papers

Despite significant advancements of deep learning-based forgery detectors for distinguishing manipulated deepfake images, most detection approaches suffer from moderate to significant performance degradation with low-quality compressed deepfake images. Because of the limited information in low-quality images, detecting low-quality deepfake remains an important challenge. In this work, we apply frequency domain learning and optimal transport theory in knowledge distillation (KD) to specifically improve the detection of low-quality compressed deepfake images. We explore transfer learning capability in KD to enable a student network to learn discriminative features from low-quality images effectively. In particular, we propose the Attention-based Deepfake detection Distiller (ADD), which consists of two novel distillations: 1) frequency attention distillation that effectively retrieves the removed high-frequency components in the student network, and 2) multi-view attention distillation that creates multiple attention vectors by slicing the teacher’s and student’s tensors under different views to transfer the teacher tensor’s distribution to the student more efficiently. Our extensive experimental results demonstrate that our approach outperforms state-of-the-art baselines in detecting low-quality compressed deepfake images.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

We study a dynamic traffic assignment model, where agents base their instantaneous routing decisions on real-time delay predictions. We formulate a mathematically concise model and derive properties of the predictors that ensure a dynamic prediction equilibrium exists. We demonstrate the versatility of our framework by showing that it subsumes the well-known full information and instantaneous information models, in addition to admitting further realistic predictors as special cases. We complement our theoretical analysis by an experimental study, in which we systematically compare the induced average travel times of different predictors, including a machine-learning model trained on data gained from previously computed equilibrium flows, both on a synthetic and a real road network.

poster session 1 @ blue 6, poster session 12 @ blue 6, poster session 1, poster session 12

3D facial landmark detection is extensively used in many research fields such as face registration, facial shape analysis, and face recognition. Most existing methods involve traditional features and 3D face models for the detection of landmarks, and their performances are limited by the hand-crafted intermediate process. In this paper, we propose a novel 3D facial landmark detection method, which directly locates the coordinates of landmarks from the 3D point cloud with a well-customized graph convolutional network. The graph convolutional network can learn geometric features adaptively for 3D facial landmark detection with the assistance of constructed 3D heatmaps, which are Gaussian functions of distances to each landmark on a 3D face. On this basis, we further develop a local surface unfolding and registration module to predict 3D landmarks from the heatmaps. We demonstrate experimentally that our method exceeds the existing approaches by a clear margin on BU-3DFE and FRGC datasets for landmark localization accuracy and stability. It also achieves high-precision results on a recent large-scale and high-quality FaceScape dataset. Generally, the proposed method forms the first baseline of the deep point cloud learning method for 3D facial landmark detection.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

In recent years, machine learning (ML) algorithms have been deployed in safety-critical and high-stake decision-making, where the fairness of algorithms is of paramount importance. Fairness in ML centers on detecting bias towards certain demographic populations induced by an ML classifier and proposes algorithmic solutions to mitigate the bias with respect to different fairness definitions. To this end, several fairness verifiers have been proposed that compute the bias in the prediction of an ML classifier—essentially beyond a finite dataset—given the probability distribution of input features. In the context of verifying linear classifiers, existing fairness verifiers are limited by accuracy due to imprecise modeling of correlations among features and scalability due to restrictive formulations of the classifiers as SSAT/SMT formulas or by sampling. In this paper, we propose an efficient fairness verifier, called FVGM, that encodes the correlations among features as a Bayesian network. In contrast to existing verifiers, FVGM proposes a stochastic subset-sum based approach for verifying linear classifiers. Experimentally, we show that FVGM leads to an accurate and scalable assessment for more diverse families of fairness-enhancing algorithms, fairness attacks, and group/causal fairness metrics than the state-of-the-art. We also demonstrate that FVGM facilitates the computation of fairness influence functions as a stepping stone to detect the source of bias induced by subsets of features.

poster session 1 @ blue 3, poster session 12 @ blue 3, oral session 12 @ blue 3, poster session 1, poster session 12, oral session 12

Some applications may require performing face recognition (FR) on third-party servers, which could be accessed by attackers with malicious intents to compromise the privacy of users’ face information. This paper advocates a practical privacy-preserving FR scheme without key management realized in the frequency domain. The new scheme first collects the components of the same frequency from different blocks of a face image to form component channels. Only part of the channels are retained and fed into the analysis network that performs an interpretable privacy-accuracy trade-off analysis to identify channels important for face image visualization but not crucial for maintaining high FR accuracy. For this purpose, the loss function of the analysis network consists of the empirical FR error loss and a face visualization penalty term, and the network is trained in an end-to-end manner. We find that with the developed analysis network, more than 94\% of the image energy can be dropped while the face recognition accuracy stays almost undegraded. In order to further protect the remaining frequency components, we propose a fast masking method. Effectiveness of the new scheme in removing the visual information of face images while maintaining their distinguishability is validated over several large face datasets. Results show that the proposed scheme achieves a recognition performance and inference time comparable to ArcFace operating on original face images directly.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Despite significant improvements on the image generation performance of Generative Adversarial Networks (GANs), generations with low visual fidelity still have been observed. As widely used metrics for GANs focus more on the overall performance of the model, evaluation on the quality of individual generations or detection of defective generations is challenging. While recent studies try to detect featuremap units that cause artifacts and evaluate individual samples, these approaches require additional resources such as external networks or a number of training data to approximate the real data manifold. In this work, we propose the concept of local activation, and devise a metric on the local activation to detect artifact generations without additional supervision. We empirically verify that our approach can detect and correct artifact generations from GANs with various datasets. Finally, we discuss a geometrical analysis to partially reveal the relation between the proposed concept and low visual fidelity.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

The challenge of talking face generation from speech lies in aligning two different modal information, audio and video, such that the mouth region corresponds to input audio. Previous methods either exploit audio-visual representation learning or leverage intermediate structural information such as landmarks and 3D models. However, they struggle to synthesize fine details of the lips varying at the phoneme level as they do not sufficiently provide visual information of the lips at the video synthesis step. To overcome this limitation, our work proposes Audio-Lip Memory that brings in visual information of the mouth region corresponding to input audio and enforces fine-grained audio-visual coherence. It stores lip motion features from sequential ground truth images in the value memory and aligns them with corresponding audio features so that they can be retrieved using audio input at inference time. Therefore, using the retrieved lip motion features as visual hints, it can easily correlate audio with visual dynamics in the synthesis step. By analyzing the memory, we demonstrate that unique lip features are stored in each memory slot at the phoneme level, capturing subtle lip motion based on memory addressing. In addition, we introduce visual-visual synchronization loss which can enhance lip-syncing performance when used along with audio-visual synchronization loss in our model. Extensive experiments are performed to verify that our method generates high-quality video with mouth shapes that best align with the input audio, outperforming previous state-of-the-art methods.

poster session 5 @ red 3, poster session 12 @ red 3, oral session 12 @ red 3, poster session 5, poster session 12, oral session 12

Active Directory is the default security management system for Windows domain networks. We study the shortest path edge interdiction problem for defending Active Directory style attack graphs. The problem is formulated as a Stackelberg game between one defender and one attacker. The attack graph contains one destination node and multiple entry nodes. The attacker's entry node is chosen by nature. The defender chooses to block a set of edges limited by his budget. The attacker then picks the shortest unblocked attack path. The defender aims to maximize the expected shortest path length for the attacker, where the expectation is taken over entry nodes. We observe that practical Active Directory attack graphs have small maximum attack path length and are structurally close to trees. We first show that even if the maximum attack path length is a constant, the problem is still $w[1]$-hard with respect to the defender's budget. Having a small maximum attack path length and a small budget is not enough to design fixed-parameter algorithms. If we further assume that the number of entry nodes is small, then we derive a fixed-parameter tractable algorithm. We then propose two other fixed-parameter algorithms by exploiting the tree-like features. One is based on tree decomposition and requires a small tree width. The other assumes a small number of splitting nodes (nodes with multiple out-going edges). Finally, the last algorithm is converted into a graph convolutional neural network based heuristic, which scales to larger graphs with more splitting nodes.

poster session 5 @ red 3, poster session 12 @ red 3, oral session 5 @ red 3, poster session 5, poster session 12, oral session 5

The expression of displeasure on a consumer’s behalf towards an organization, product, or event is denoted via the speech act known as complaint. Customers typically post reviews on retail websites and various social media platforms about the products or services they purchase, and the reviews may include complaints about the products or services. Automatic detection of consumers' complaints about items or services they buy can be critical for organizations and online merchants. Since they can use this insight to meet the customers' requirements, including handling and addressing the complaints. Previous studies on Complaint Identification (CI) are limited to text. Images posted with the reviews can provide cues to better identify complaints, thus emphasizing the importance of incorporating multi-modal inputs into the process. Furthermore, the customer's emotional state has a significant impact on the complaint expression, since any speech act is generally influenced by emotions. As a result, the impact of emotion and sentiment on automatic complaint identification must also be investigated. One of the major contributions of this work is a new dataset- Complaint, Emotion, and Sentiment Annotated Multi-modal Amazon Reviews Dataset (CESAMARD) which is a collection of opinionated texts (reviews) and images of the products posted on the website of the retail giant Amazon. We present an attention-based multi-modal, adversarial multi-task deep neural network model for complaint detection to demonstrate the utility of the multi-modal dataset. Experimental results indicate that the multi-modality and multi-tasking complaint identification outperforms uni-modal and single-task variants.

poster session 3 @ red 6, poster session 12 @ red 6, poster session 3, poster session 12

DEtection TRansformer (DETR) is a recently proposed method that streamlines the detection pipeline and achieves competitive results against two-stage detectors such as Faster-RCNN. The DETR models get rid of complex anchor generation and post-processing procedures thereby making the detection pipeline more intuitive. However, the numerous redundant parameters in transformers make the DETR models computation and storage intensive, which seriously hinder them to be deployed on the resources-constrained devices. In this paper, to obtain a compact end-to-end detection framework, we propose to deeply compress the transformers with low-rank tensor decomposition. The basic idea of the tensor-based compression is to represent the large-scale weight matrix in one network layer with a chain of low-order matrices. Furthermore, we propose a gated multi-head attention (GMHA) module to mitigate the accuracy drop of the tensor-compressed DETR models. In GMHA, each attention head has an independent gate to determine the passed attention value. The redundant attention information can be suppressed by adopting the normalized gates. Lastly, to obtain fully compressed DETR models, a low-bitwidth quantization technique is introduced for further reducing the model storage size. Based on the proposed methods, we can achieve significant parameter and model size reduction while maintaining high detection performance. We conduct extensive experiments on the COCO dataset to validate the effectiveness of our tensor-compressed (tensorized) DETR models. The experimental results show that we can attain 3.7 times full model compression with 482 times feed forward network (FFN) parameter reduction and only 0.6 points accuracy drop.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

We present an extremely simple Ultra-Resolution Style Transfer framework, termed URST, to flexibly process arbitrary high-resolution images (e.g., 10000x10000 pixels) style transfer for the first time. Most of the existing state-of-the-art methods would fall short due to massive memory cost and small stroke size when processing ultra-high resolution images. URST completely avoids the memory problem caused by ultra-high resolution images by (1) dividing the image into small patches and (2) performing patch-wise style transfer with a novel Thumbnail Instance Normalization (TIN). Specifically, TIN can extract thumbnail features' normalization statistics and apply them to small patches, ensuring the style consistency among different patches. Overall, the URST framework has three merits compared to prior arts. (1) We divide input image into small patches and adopt TIN, successfully transferring image style with arbitrary high-resolution. (2) Experiments show that our URST surpasses existing SOTA methods on ultra-high resolution images benefiting from the effectiveness of the proposed stroke perceptual loss in enlarging the stroke size. (3) Our URST can be easily plugged into most existing style transfer methods and directly improve their performance even without training. Code is available at https://git.io/URST.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

Cross-lingual pre-training has achieved great successes using monolingual and bilingual plain text corpora. However, most pre-trained models neglect multilingual knowledge, which is language agnostic but comprises abundant cross-lingual structure alignment. In this paper, we propose XLM-K, a cross-lingual language model incorporating multilingual knowledge in pre-training. XLM-K augments existing multilingual pre-training with two knowledge tasks, namely Masked Entity Prediction Task and Object Entailment Task. We evaluate XLM-K on MLQA, NER and XNLI. Experimental results clearly demonstrate significant improvements over existing multilingual language models. The results on MLQA and NER exhibit the superiority of XLM-K in knowledge related tasks. The success in XNLI shows a better cross-lingual transferability obtained in XLM-K. What is more, we provide a detailed probing analysis to confirm the desired knowledge captured in our pre-training regimen.

poster session 5 @ red 5, poster session 12 @ red 5, poster session 5, poster session 12

This paper proposes a novel two-stage hypergraph-based framework, dubbed ADaptive Hypergraph Neural Network (AD-HNN) to estimate multiple human poses from a single image, with a keypoint localization network and an Adaptive-Pose Hypergraph Neural Network (AP-HNN) added onto the former network. For providing better guided representations of AP-HNN, we employ a Semantic Interaction Convolution (SIC) module within the initial localization network to acquire more explicit predictions. Build upon this, we design a novel adaptive hypergraph to represent a human body for capturing high-order semantic relations among different joints. Notably, it can adaptively adjust the relations between joints and seek the most reasonable structure for the variable poses to benefit the keypoint localization. These two stages are combined to be trained in an end-to-end fashion. Unlike traditional Graph Convolutional Networks (GCNs) that are based on a fixed tree structure, AP-HNN can deal with ambiguity in human pose estimation. Experimental results demonstrate that the AD-HNN achieves state-of-the-art performance both on the MS-COCO, MPII and CrowdPose datasets.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

N-ary relational knowledge base (KBs) embedding aims to map binary and beyond-binary facts into low-dimensional vector space simultaneously. Existing approaches typically decompose n-ary relational facts into subtuples (entity pairs, triples or quintuples, etc.), and they generally model n-ary relational KBs in Euclidean space. However, n-ary relational facts are semantically and structurally intact, decomposition leads to the loss of global information and undermines the semantical and structural integrity. Moreover, compared to the binary relational KBs, n-ary ones are characterized by more abundant and complicated hierarchy structures, which could not be well expressed in Euclidean space. To address the issues, we propose a gyro-polygon embedding approach to realize n-ary fact integrity keeping and hierarchy capturing, termed as PolygonE. Specifically, n-ary relational facts are modeled as gyro-polygons in the hyperbolic space, where we denote entities in facts as vertexes of gyro-polygons and relations as entity translocation operations. Importantly, we design a fact plausibility measuring strategy based on the vertex-gyrocentroid geodesic to optimize the relation-adjusted gyro-polygon. Extensive experiments demonstrate that PolygonE shows SOTA performance on all benchmark datasets, generalizability to binary data, and applicability to arbitrary arity fact. Finally, we also visualize the embedding to help comprehend PolygonE's awareness of hierarchies.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

On-device directional hearing requires audio source separation from a given direction while achieving stringent human-imperceptible latency requirements. While neural nets can achieve significantly better performance than traditional beamformers, all existing models fall short of supporting low-latency causal inference on computationally-constrained wearables. We present DeepBeam, a hybrid model that combines traditional beamformers with a custom lightweight neural net. The former reduces the computational burden of the latter and also improves its generalizability, while the latter is designed to further reduce the memory and computational overhead to enable real-time and low-latency operations. Our evaluation shows comparable performance to state-of-the-art causal inference models on synthetic data while achieving a 5x reduction of model size, 4x reduction of computation per second, 5x reduction in processing time and generalizing better to real hardware data. Further, our real-time hybrid model runs in 8 ms on mobile CPUs designed for low-power wearable devices and achieves an end-to-end latency of 17.5 ms.

poster session 5 @ red 5, poster session 12 @ red 5, poster session 5, poster session 12

With various facial manipulation techniques arising, face forgery detection has drawn growing attention due to security concerns. Previous works always formulate face forgery detection as a classification problem based on cross-entropy loss, which emphasizes category-level differences rather than the essential discrepancies between real and fake faces, limiting model generalization in unseen domains. To address this issue, we propose a novel face forgery detection framework, named Dual Contrastive Learning (DCL), which specially constructs positive and negative paired data and performs designed contrastive learning at different granularities to learn generalized feature representation. Concretely, combined with the hard sample selection strategy, Inter-Instance Contrastive Learning (Inter-ICL) is first proposed to promote task-related discriminative features learning by especially constructing instance pairs. Moreover, to further explore the essential discrepancies, Intra-Instance Contrastive Learning (Intra-ICL) is introduced to focus on the local content inconsistencies prevalent in the forged faces by constructing local-region pairs inside instances. Extensive experiments and visualizations on several datasets demonstrate the generalization of our method against the state-of-the-art competitors.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Heuristics for lifted planning are not yet as informed as the best heuristics for ground planning. Recent work introduced the idea of using Datalog programs to compute the additive heuristic over lifted tasks. Based on this work, we show how to compute the more informed FF heuristic in a lifted manner. We extend the Datalog program with executable annotations that can also be used to define other delete-relaxation heuristics. In our experiments, we show that a planner using the lifted FF implementation produces state-of-the-art results for lifted planners. It also reduces the gap to state-of-the-art ground planners in domains where grounding is feasible.

poster session 3 @ blue 4, poster session 12 @ blue 4, poster session 3, poster session 12

In this work, we focus on a very practical problem: image segmentation under rain conditions. Image deraining is a classic low-level restoration task, while image segmentation is a typical high-level understanding task. Most of the existing methods intuitively employ the bottom-up paradigm by taking deraining as a preprocessing step for subsequent segmentation. However, our statistical analysis indicates that not only deraining would benefit segmentation (bottom-up), but also segmentation would further improve deraining performance (top-down) in turn. This motivates us to solve the rainy image segmentation task within a novel top-down and bottom-up unified paradigm, in which two sub-tasks are alternatively performed and collaborated with each other. Specifically, the bottom-up procedure yields both clearer images and rain-robust features from both image and feature domains, so as to ease the segmentation ambiguity caused by rain streaks. The top-down procedure adopts semantics to adaptively guide the restoration for different contents via a novel multi-path semantic attentive module (SAM). Thus the deraining and segmentation could boost the performance of each other cooperatively and progressively. Extensive experiments and ablations demonstrate that the proposed method outperforms the state-of-the-art on rainy image segmentation.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

A 6-hour early detection of sepsis leads to a significant increase in the chance of surviving it. Previous sepsis early detection studies have focused on improving the performance of supervised learning algorithms while ignoring the potential correlation in data mining, and there was no reliable method to deal with the problem of incomplete data. In this paper, we proposed the Denoising Transformer AutoEncoder (DTAE) for the first time combining transformer and unsupervised learning. DTAE can learn the correlation of the features required for early detection of sepsis without the label. This method can effectively solve the problems of data sparsity and noise and discover the potential correlation of features by adding DTAE enhancement module without modifying the existing algorithms. Finally, the experimental results show that the proposed method improves the existing algorithms and achieves the best results of early detection.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

Recently, deep unfolding networks (DUNs) based on optimization algorithms have received increasing attention, and their high efficiency has been confirmed by many experimental and theoretical results. Since this type of networks combines model-based traditional optimization algorithms, they have high interpretability. In addition, ordinary differential equations (ODEs) are often used to explain deep neural networks, and provide some inspiration for designing innovative network models. In this paper, we transform DUNs into first-order ODE forms, and propose a high-order numerical architecture for ODE-inspired deep unfolding networks. To the best of our knowledge, this is the first work to establish the relationship between DUNs and ODEs. Moreover, we take two representative DUNs as examples, apply our architecture to them and design novel DUNs. In theory, we prove the existence, uniqueness of the solution and convergence of the proposed network, and also prove that our network obtains a fast linear convergence rate. Extensive experiments verify the effectiveness and advantages of our architecture.

poster session 2 @ red 1, poster session 12 @ red 1, oral session 12 @ red 1, poster session 2, poster session 12, oral session 12

Despite the super-human accuracy of recent deep models in NLP tasks, their robustness is reportedly limited due to their reliance on spurious patterns. We thus aim to leverage contrastive learning and counterfactual augmentation for robustness. For augmentation, existing work either requires humans to add counterfactuals to the dataset or machines to automatically matches near-counterfactuals already in the dataset. Unlike existing augmentation is affected by spurious correlations, ours, by synthesizing “a set” of counterfactuals, and making a collective decision on the distribution of predictions on this set, can robustly supervise the causality of each term. Our empirical results show that our approach, by collective decisions, is less sensitive to task model bias of attribution-based synthesis, and thus achieves significant improvements, in diverse dimensions: 1) counterfactual robustness, 2) cross-domain generalization, and 3) generalization from scarce data.

poster session 3 @ red 4, poster session 12 @ red 4, poster session 3, poster session 12

Deepfake has ignited hot research interests in both academia and industry due to its potential security threats. Many countermeasures have been proposed to mitigate such risks. Current Deepfake detection methods achieve superior performances in dealing with low-visual-quality Deepfake media which can be distinguished by the obvious visual artifacts. However, with the development of deep generative models, the realism of Deepfake media has been significantly improved and becomes tough challenging to current detection models. In this paper, we propose a frame inference-based detection framework (FInfer) to solve the problem of high-visual-quality Deepfake detection. Specifically, we first learn the referenced representations of the current and future frames’ faces. Then, the current frames’ facial representations are utilized to predict the future frames’ facial representations by using an autoregressive model. Finally, a representation-prediction loss is devised to maximize the discriminability of real videos and fake videos. We demonstrate the effectiveness of our FInfer framework through information theory analyses. The entropy and mutual information analyses indicate the correlation between the predicted representations and referenced representations in real videos is higher than that of high-visual-quality Deepfake videos. Extensive experiments demonstrate the performance of our method is promising in terms of in-dataset detection performance, detection efficiency, and cross-dataset detection performance in high-visual-quality Deepfake videos.

poster session 6 @ blue 2, poster session 12 @ blue 2, oral session 12 @ blue 2, poster session 6, poster session 12, oral session 12

Convolutional Neural Networks (CNNs) perform very well in image classification and object detection in recent years, but even the most advanced models have limited rotation invariance. Known solutions include the enhancement of training data and the increase of rotation invariance by globally merging the rotation equivariant features. These methods either increase the workload of training or increase the number of model parameters. To address this problem, this paper proposes a module that can be inserted into the existing networks, and directly incorporates the rotation invariance into the feature extraction layers of the CNNs. This module does not have learnable parameters and will not increase the complexity of the model. At the same time, only by training the upright data, it can perform well on the rotated testing set. These ad-vantages will be suitable for fields such as biomedicine and astronomy where it is difficult to obtain upright samples or the target has no directionality. Evaluate our module with LeNet-5, ResNet-18 and tiny-yolov3, we get impressive results.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

Generally, humans are more skilled at perceiving differences between high-quality (HQ) and low-quality (LQ) images than directly judging the quality of a single LQ image. This situation also applies to image quality assessment (IQA). Although recent no-reference (NR-IQA) methods have made great progress to predict image quality free from the reference image, they still have the potential to achieve better performance since HQ image information is not fully exploited. In contrast, full-reference (FR-IQA) methods tend to provide more reliable quality evaluation, but its practicability is affected by the requirement for pixel-level aligned reference images. To address this, we firstly propose the content-variant reference method via knowledge distillation (CVRKD-IQA). Specifically, we use non-aligned reference (NAR) images to introduce various prior distributions of high-quality images. The comparisons of distribution differences between HQ and LQ images can help our model better assess the image quality. Further, the knowledge distillation transfers more HQ-LQ distribution difference information from the FR-teacher to the NAR-student and stabilizing CVRKD-IQA performance. Moreover, to fully mine the local-global combined information, while achieving faster inference speed, our model directly processes multiple image patches from the input with the MLP-mixer. Cross-dataset experiments verify that our model can outperform all NAR/NR-IQA SOTAs, even reach comparable performance than FR-IQA methods on some occasions. Since the content-variant and non-aligned reference HQ images are easy to obtain, our model can support more IQA applications with its robustness to content variations. Our code is available: https://github.com/guanghaoyin/CVRKD-IQA.

poster session 5 @ red 2, poster session 12 @ red 2, oral session 12 @ red 2, poster session 5, poster session 12, oral session 12

Recent non-local self-attention methods have proven to be effective in capturing long-range dependencies for semantic segmentation. These methods usually form a similarity map of \(\mathbb{R}^{C\times C}\) (by compressing spatial dimensions) or \(\mathbb{R}^{HW\times HW}\) (by compressing channels) to describe the feature relations along either channel or spatial dimensions, where \(C\) is the number of channels, \(H\) and \(W\) are the spatial dimensions of the input feature map. However, such practices tend to condense feature dependencies along the other dimensions, hence causing attention missing, which might lead to inferior results for small/thin categories or inconsistent segmentation inside large objects. To address this problem, we propose a new approach, namely Fully Attentional Network (FLANet), to encode both spatial and channel attentions in a single similarity map while maintaining high computational efficiency. Specifically, for each channel map, our FLANet can harvest feature responses from all other channel maps, and the associated spatial positions as well, through a novel fully attentional module. Our new method has achieved state-of-the-art performance on three challenging semantic segmentation datasets, i.e., 83.6\%, 46.99\%, and 88.5\% on the Cityscapes test set, the ADE20K validation set, and the PASCAL VOC test set, respectively.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

Detection of manipulated face images has attracted a lot of interest recently. Various schemes have been proposed to tackle this challenging problem, where the patch-based approaches are shown to be promising. However, the existing patch-based approaches tend to treat different patches equally, which do not fully exploit the patch discrepancy for effective feature learning. In this paper, we propose a Patch Diffusion (PD) module which can be integrated into the existing face manipulation detection networks to boost the performance. The PD consists of Discrepancy Patch Feature Learning (DPFL) and Attention-Aware Message Passing (AMP). The DPFL effectively learns the patch features by a newly designed Pairwise Patch Loss (PPLoss), which takes both the patch importance and correlations into consideration. The AMP diffuses the patches through attention-aware message passing in a graph network, where the attentions are explicitly computed based on the patch features learnt in DPFL. We integrate our PD module into four recent face manipulation detection networks, and carry out the experiments on four popular datasets. The results demonstrate that our PD module is able to boost the performance of the existing networks for face manipulation detection.

poster session 6 @ blue 2, poster session 12 @ blue 2, oral session 6 @ blue 2, poster session 6, poster session 12, oral session 6

Out-of-Distribution (OOD) detection is essential in real-world applications, which has attracted increasing attention in recent years. However, most existing OOD detection methods require many labeled In-Distribution (ID) data, causing a heavy labeling cost. In this paper, we focus on the more realistic scenario, where limited labeled data and abundant unlabeled data are available, and these unlabeled data are mixed with ID and OOD samples. We propose the Adaptive In-Out-aware Learning (AIOL) method, in which we employ the appropriate temperature to adaptively select potential ID and OOD samples from the mixed unlabeled data and consider the entropy over them for OOD detection. Moreover, since the test data in realistic applications may contain OOD samples whose classes are not in the mixed unlabeled data (we call them unseen OOD classes), data augmentation techniques are brought into the method to further improve the performance. The experiments are conducted on various benchmark datasets, which demonstrate the superiority of our method.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

With the rapid development of facial forgery techniques, forgery detection has attracted more and more attention due to security concerns. Existing approaches attempt to use frequency information to mine subtle artifacts under high-quality forged faces. However, the exploitation of frequency information is coarse-grained, and more importantly, their vanilla learning process struggles to extract fine-grained forgery traces. To address this issue, we propose a progressive enhancement learning framework to exploit both the RGB and fine-grained frequency clues. Specifically, we perform a fine-grained decomposition of RGB images to completely decouple the real and fake traces in the frequency space. Subsequently, we propose a progressive enhancement learning framework based on a two-branch network, combined with self-enhancement and mutual-enhancement modules. The self-enhancement module captures the traces in different input spaces based on spatial noise enhancement and channel attention. The Mutual-enhancement module concurrently enhances RGB and frequency features by communicating in the shared spatial dimension. The progressive enhancement process facilitates the learning of discriminative features with fine-grained face forgery clues. Extensive experiments on several datasets show that our method outperforms the state-of-the-art face forgery detection methods.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Nowadays, due to the breakthrough in natural language generation (NLG), including machine translation, document summarization, image captioning, etc NLG models have been encapsulated in cloud APIs to serve over half a billion people worldwide and process over one hundred billion word generations per day. Thus, NLG APIs have already become essential profitable services in many commercial companies. Due to the substantial financial and intellectual investments, service providers adopt a pay-as-you-use policy to promote sustainable market growth. However, recent works have shown that cloud platforms suffer from financial losses imposed by model extraction attacks, which aim to imitate the functionality and utility of the victim services, thus violating the intellectual property (IP) of cloud APIs. This work targets at protecting IP of NLG APIs by identifying the attackers who have utilized watermarked responses from the victim NLG APIs. However, most existing watermarking techniques are not directly amenable for IP protection of NLG APIs. To bridge this gap, we first present a novel watermarking method for text generation APIs by conducting lexical modification to the original outputs. Compared with the competitive baselines, our watermark approach achieves better identifiable performance in terms of p-value, with fewer semantic losses. In addition, our watermarks are more understandable and intuitive to humans than the baselines. Finally, the empirical studies show our approach is also applicable to queries from different domains, and is effective on the attacker trained on a mixture of the corpus which includes less than 10% watermarked samples.

poster session 5 @ red 5, poster session 12 @ red 5, oral session 12 @ red 5, poster session 5, poster session 12, oral session 12

Bias mitigation in machine learning models is imperative, yet challenging. While several approaches have been proposed, one view towards mitigating bias is through adversarial learning. A discriminator is used to identify the bias attributes such as gender, age or race in question. This discriminator is used adversarially to ensure that it cannot distinguish the bias attributes. The main drawback in such a model is that it directly introduces a trade-off with accuracy as the features that the discriminator deems to be sensitive for discrimination of bias could be correlated with classification. In this work we solve the problem. We show that a biased discriminator can actually be used to improve this bias-accuracy tradeoff. Specifically, this is achieved by using a feature masking approach using the discriminator's gradients. We ensure that the features favoured for the bias discrimination are de-emphasized and the unbiased features are enhanced during classification. We show that this simple approach works well to reduce bias as well as improve accuracy significantly. We evaluate the proposed model on standard benchmarks. We improve the accuracy of the adversarial methods while maintaining or even improving the unbiasness and also outperform several other recent methods.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

RGB-infrared person re-identification is an emerging cross-modality re-identification task, which is very challenging due to significant modality discrepancy between RGB and infrared images. In this work, we propose a novel modality-adaptive mixup and invariant decomposition (MID) approach for RGB-infrared person re-identification towards learning modality-invariant and discriminative representations. MID designs a modality-adaptive mixup scheme to generate suitable mixed modality images between RGB and infrared images for mitigating the inherent modality discrepancy at the pixel-level. It formulates modality mixup procedure as Markov decision process, where an actor-critic agent learns dynamical and local linear interpolation policy between different regions of cross-modality images under a deep reinforcement learning framework. Such policy guarantees modality-invariance in a more continuous latent space and avoids manifold intrusion by the corrupted mixed modality samples. Moreover, to further counter modality discrepancy and enforce invariant visual semantics at the feature-level, MID employs modality-adaptive convolution decomposition to disassemble a regular convolution layer into modality-specific basis layers and a modality-shared coefficient layer. Extensive experimental results on two challenging benchmarks demonstrate superior performance of MID over state-of-the-art methods.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

In this paper, we propose TransMEF: a transformer-based multi-exposure image fusion framework via self-supervised multi-task learning. The framework is based on an encoder-decoder network, which can be trained on large natural datasets without the need of ground truth fusion images. We design three self-supervised reconstruction tasks according to the characteristics of multi-exposure images and conduct the three tasks simultaneously through multi-task learning, so that the network can learn the characteristics of multi-exposure images and extract more generalized features. In addition, to compensate for the defect in establishing long-range dependencies in CNN-based architectures, we design an encoder that combines a CNN module with a Transformer module to enable the network to focus on both local and global information. We evaluated our method and compared it to 11 competitive traditional and deep learning-based methods on a latest released multi-exposure image fusion benchmark dataset, and our method achieves the best performance in both subjective and objective evaluations.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

We present a simple and concise semantics for temporal planning. Our semantics are developed and formalised in the logic of the interactive theorem prover Isabelle/HOL. We derive from those semantics a validation algorithm for temporal planning and show, using a formal proof in Isabelle/HOL, that this validation algorithm implements our semantics. We experimentally evaluate our verified validation algorithm and show that it performs reasonably well.

poster session 3 @ blue 4, poster session 12 @ blue 4, poster session 3, poster session 12

Zero-shot learning (ZSL) aims to recognize novel classes by transferring semantic knowledge from seen classes to unseen ones. Semantic knowledge is learned from attribute descriptions shared between different classes, which are strong prior for localization of object attribute for representing discriminative region features enabling significant visual-semantic interaction. Although few attention-based models have attempted to learn such region features in a single image, the transferability and discriminative attribute localization of visual features are typically neglected. In this paper, we propose an attribute-guided Transformer network to learn the attribute localization for discriminative visual-semantic embedding representations in ZSL, termed TransZero. Specifically, TransZero takes a feature augmentation encoder to alleviate the cross-dataset bias between ImageNet and ZSL benchmarks and improve the transferability of visual features by reducing the entangled relative geometry relationships among region features. To learn locality-augmented visual features, TransZero employs a visual-semantic decoder to localize the most relevant image regions to each attributes from a given image under the guidance of attribute semantic information. Then, the locality-augmented visual features and semantic vectors are used for conducting effective visual-semantic interaction in a visual-semantic embedding network. Extensive experiments show that TransZero achieves a new state-of-the-art on three ZSL benchmarks. The codes are available at: \url{https://github.com/shiming-chen/TransZero}.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

Recently, numerous algorithms have been developed to tackle the problem of light field super-resolution (LFSR), i.e., super-resolving low-resolution light fields to gain high-resolution views. Despite delivering encouraging results, these approaches are all convolution-based, and are naturally weak in global relation modeling of sub-aperture images necessarily to characterize the inherent structure of light fields. In this paper, we put forth a novel formulation built upon Transformers, by treating LFSR as a sequence-to-sequence reconstruction task. In particular, our model regards sub-aperture images of each vertical or horizontal angular view as a sequence, and establishes long-range geometric dependencies within each sequence via a spatial-angular locally-enhanced self-attention layer, which maintains the locality of each sub-aperture image as well. Additionally, to better recover image details, we propose a detail-preserving Transformer (termed as DPT), by leveraging gradient maps of light field to guide the sequence learning. DPT consists of two branches, with each associated with a Transformer for learning from an original or gradient image sequence. The two branches are finally fused to obtain comprehensive feature representations for reconstruction. Evaluations are conducted on a number of light field datasets, including real-world scenes and synthetic data. The proposed method achieves superior performance comparing with other state-of-the-art schemes. Our code is publicly available at: {https://github.com/BITszwang/DPT}.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

Deep learning has made remarkable achievements for single image haze removal. However, existing deep dehazing models only give deterministic results without discussing the uncertainty of them. There exist two types of uncertainty in the dehazing models: aleatoric uncertainty that comes from noise inherent in the observations and epistemic uncertainty that accounts for uncertainty in the model. In this paper, we propose a novel uncertainty-driven dehazing network (UDN) that improves the dehazing results by exploiting the relationship between the uncertain and confident representations. We first introduce an Uncertainty Estimation Block (UEB) to predict the aleatoric and epistemic uncertainty together. Then, we propose an Uncertainty-aware Feature Modulation (UFM) block to adaptively enhance the learned features. UFM predicts a convolution kernel and channel-wise modulation cofficients conitioned on the uncertainty weighted representation. Moreover, we develop an uncertainty-driven self-distillation loss to improve the uncertain representation by transferring the knowledge from the confident one. Extensive experimental results on synthetic datasets and real-world images show that UDN achieves significant quantitative and qualitative improvements, outperforming the state-of-the-arts.

poster session 5 @ red 2, poster session 12 @ red 2, oral session 5 @ red 2, poster session 5, poster session 12, oral session 5

While deep neural network-based video denoising methods have achieved promising results, it is still hard to deploy them on mobile devices due to their high computational cost and memory demands. This paper aims to develop a lightweight deep video denoising method that is friendly to resource-constrained mobile devices. Inspired by the facts that 1) consecutive video frames usually contain redundant temporal coherency, and 2) neural networks are usually over-parameterized, we propose a multi-input multi-output (MIMO) paradigm to process consecutive video frames within one-forward-pass. The basic idea is concretized to a novel architecture termed Recurrent Multi-output Network (ReMoNet), which consists of recurrent temporal fusion and temporal aggregation blocks and is further reinforced by similarity-based mutual distillation. We conduct extensive experiments on NVIDIA GPU and Qualcomm Snapdragon 888 mobile platform with Gaussian noise and simulated Image-Signal-Processor (ISP) noise. The experimental results show that ReMoNet is both effective and efficient on video denoising. Moreover, we show that ReMoNet is more robust under higher noise level scenarios.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

Even though graphic layout generation has attracted growing attention recently, it is still challenging to synthesis realistic and diverse layouts, due to the complicated element relationships and varied element arrangements. In this work, we seek to improve the performance of layout generation by incorporating the concept of regions, which consist of a smaller number of elements and appears like a simple layout, into the generation process. Specifically, we leverage Variational Autoencoder (VAE) as the overall architecture and decompose the decoding process into two stages. The first stage predicts representations for regions, and the second stage fills in the detailed position for each element within the region based on the predicted region representation. Compared to prior studies that merely abstract the layout into a list of elements and generate all the element positions in one go, our approach has at least two advantages. First, by the two-stage decoding, our approach decouples the complex layout generation task into several simple layout generation tasks, which reduces the problem difficulty. Second, the predicted regions can help the model roughly know what the graphic layout looks like and serve as global context to improve the generation of detailed element positions. Qualitative and quantitative experiments demonstrate that our approach significantly outperforms the existing methods, especially on the complex graphic layouts.

poster session 6 @ blue 2, poster session 12 @ blue 2, oral session 12 @ blue 2, poster session 6, poster session 12, oral session 12

The task of Legal Statute Identification (LSI) aims to identify the legal statutes that are relevant to a given description of facts or evidence of a legal case. Existing methods only utilize the textual content of facts and legal articles to guide such a task. However, the citation network among case documents and legal statutes is a rich source of additional information, which is not considered by existing models. In this work, we take the first step towards utilising both the text and the legal citation network for the LSI task. We curate a large novel dataset for this task, including facts of cases from several major Indian Courts of Law, and statutes from the Indian Penal Code (IPC). Modeling the statutes and training documents as a heterogeneous graph, our proposed model LeSICiN can learn rich textual and graphical features, and can also tune itself to correlate these features. Thereafter, the model can be used to inductively predict links between test documents (new nodes whose graphical features are not available to the model) and statutes (existing nodes). Extensive experiments on the dataset show that our model comfortably outperforms several state-of-the-art baselines, by exploiting the graphical structure along with textual features.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

We introduce VAST, the Valence-Assessing Semantics Test, a novel intrinsic evaluation task for contextualized word embeddings (CWEs). Despite the widespread use of contextualizing language models (LMs), researchers have no intrinsic evaluation task for understanding the semantic quality of CWEs and their unique properties as related to contextualization, the change in the vector representation of a word based on surrounding words; tokenization, the breaking of uncommon words into subcomponents; and LM-specific geometry learned during training. VAST uses valence, the association of a word with pleasantness, to measure the correspondence of word-level LM semantics with widely used human judgments, and examines the effects of contextualization, tokenization, and LM-specific geometry. Because prior research has found that CWEs from OpenAI's 2019 English-language causal LM GPT-2 perform poorly on other intrinsic evaluations, we select GPT-2 as our primary subject, and include results showing that VAST is useful for 7 other LMs, and can be used in 7 languages. GPT-2 results show that the semantics of a word are more similar to the semantics of context in layers closer to model output, such that VAST scores diverge between our contextual settings, ranging from Pearson’s rho of .55 to .77 in layer 11. We also show that multiply tokenized words are not semantically encoded until layer 8, where they achieve Pearson’s rho of .46, indicating the presence of an encoding process for multiply tokenized words which differs from that of singly tokenized words, for which rho is highest in layer 0. We find that a few neurons with values having greater magnitude than the rest mask word-level semantics in GPT-2’s top layer, but that word-level semantics can be recovered by nullifying non-semantic principal components: Pearson’s rho in the top layer improves from .32 to .76. Downstream POS tagging and sentence classification experiments indicate that the GPT-2 uses these principal components for non-semantic purposes, such as to represent sentence-level syntax relevant to next-word prediction. After isolating semantics, we show the utility of VAST for understanding LM semantics via improvements over related work on four word similarity tasks, with a score of .50 on SimLex-999, better than the previous best of .45 for GPT-2. Finally, we show that 8 of 10 WEAT bias tests, which compare differences in word embedding associations between groups of words, exhibit more stereotype-congruent biases after isolating semantics, indicating that non-semantic structures in LMs also mask social biases.

poster session 3 @ red 4, poster session 12 @ red 4, poster session 3, poster session 12

Most approaches for goal recognition rely on specifications of the possible dynamics of the actor in the environment when pursuing a goal. These specifications suffer from two key issues. First, encoding these dynamics requires careful design by a domain expert, which is often not robust to noise at recognition time. Second, existing approaches often need costly real-time computations to reason about the likelihood of each potential goal. In this paper, we develop a framework that combines model-free reinforcement learning and goal recognition to alleviate the need for careful, manual domain design, and the need for costly online executions. This framework consists of two main stages: Offline learning of policies or utility functions for each potential goal, and online inference. We provide a first instance of this framework using tabular Q-learning for the learning stage, as well as three measures that can be used to perform the inference stage. The resulting instantiation achieves state-of-the-art performance against goal recognizers on standard evaluation domains and superior performance in noisy environments.

poster session 3 @ blue 4, poster session 12 @ blue 4, poster session 3, poster session 12

Despite the advances in multi-person pose estimation, state-of-the-art techniques only deliver the human pose structure.Yet, they do not leverage the keypoints of human pose to deliver whole-body shape information for human instance segmentation. This paper presents PosePlusSeg, a joint model designed for both human pose estimation and instance segmentation. For pose estimation, PosePlusSeg first takes a bottom-up approach to detect the soft and hard keypoints of individuals by producing a strong keypoint heat map, then improves the keypoint detection confidence score by producing a body heat map. For instance segmentation, PosePlusSeg generates a mask offset where keypoint is defined as a centroid for the pixels in the embedding space, enabling instance-level segmentation for the human class. Finally, we propose a new pose and instance segmentation algorithm that enables PosePlusSeg to determine the joint structure of the human pose and instance segmentation. Experiments using the COCO challenging dataset demonstrate that PosePlusSeg copes better with challenging scenarios, like occlusions, en-tangled limbs, and overlapped people. PosePlusSeg outperforms state-of-the-art detection-based approaches achieving a 0.728 mAP for human pose estimation and a 0.445 mAP for instance segmentation. Code has been made available at: https://github.com/RaiseLab/PosePlusSeg.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

We consider the problem of maximizing the Nash social welfare when allocating a set $\calG$ of indivisible goods to a set $\calN$ of agents. We study instances, in which all agents have 2-value additive valuations: The value of every agent $i \in \calN$ for every good $j \in \calG$ is $v_{ij} \in \{p,q\}$, for $p,q \in \NN$, $p \le q$. In this work, we design an algorithm to compute an optimal allocation in polynomial time if $p$ divides $q$, i.e., when $p=1$ and $q \in \NN$ after appropriate scaling. The problem is \classNP-hard whenever $p$ and $q$ are coprime and $p \ge 3$. In terms of approximation, we present positive and negative results for general $p$ and $q$. We show that our algorithm obtains an approximation ratio of at most 1.0345. Moreover, we prove that the problem is \classAPX-hard, with a lower bound of $1.000015$ achieved at $p/q = 4/5$.

poster session 1 @ blue 6, poster session 12 @ blue 6, oral session 12 @ blue 6, poster session 1, poster session 12, oral session 12

Constrained optimization problems arise frequently in classical machine learning. There exist frameworks addressing constrained optimization, for instance, CVXPY, but so far, in contrast to deep learning frameworks, GPU support is limited. Here, we provide a framework for solving constrained optimization problems on the GPU. The framework allows the user to specify constrained optimization problems in an easy-to-read modeling language. A solver is then automatically generated from this specification. When run on the GPU, the solver outperforms state-of-the-art approaches like CVXPY combined with a GPU-accelerated solver such as cuOSQP or SCS by several orders of magnitude.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

The proliferation of fake news has grown into a global concern with adverse socio-political and economical impact. In recent years, machine learning has emerged as a promising approach to the automation of detecting and tracking fake news at scale. Current state of the art in the identification of fake news is generally focused on semantic analysis of the text, resulting in promising performance in automated detection of fake news. However, fake news campaigns are also evolving in response to such new technologies by mimicking semantic features of genuine news, which can significantly affect the performance of fake news classifiers trained on contextually limited features. In this work, we propose a novel hybrid deep learning model for fake news detection that augments the semantic characteristics of the news with features extracted from the structure of the dissemination network. To this end, we first extend the LIAR dataset by integrating sentiment and affective features to the data, and then use a BERT-based model to obtain a representation of the text. Moreover, we propose a novel approach for fake news detection based on Graph Attention Networks to leverage the user-centric features and graph features of news residing social network in addition to the features extracted in the previous steps. Experimental evaluation of our approach shows classification accuracy of 97% on the Politifact dataset. We also examined the generalizability of our proposed model on the BuzzFeed dataset, resulting in an accuracy 89.50%.

poster session 5 @ blue 4, poster session 12 @ blue 4, poster session 5, poster session 12

Low-light image enhancement aims to improve an image's visibility while keeping its visual naturalness. Different from existing methods, which tend to accomplish the relighting task directly, we investigate the intrinsic degradation and relight the low-light image while refining the details and color in two steps. Inspired by the color image formulation (diffuse illumination color plus environment illumination color), we first estimate the degradation from low-light inputs to simulate the distortion of environment illumination color, and then refine the content to recover the loss of diffuse illumination color. To this end, we propose a novel Degradation-to-Refinement Generation Network (DRGN). Its distinctive features can be summarized as 1) A novel two-step generation network for degradation learning and content refinement. It is not only superior to one-step methods, but also capable of synthesizing sufficient paired samples to benefit the model training; 2) A multi-resolution fusion network to represent the target information (degradation or contents) in a multi-scale cooperative manner, which is more effective to address the complex unmixing problems. Extensive experiments on both the enhancement task and the joint detection task have verified the effectiveness and efficiency of our proposed method, surpassing the SOTA by 1.59dB on average and 3.18\% in mAP on the ExDark dataset. The code will be available soon.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

We consider a novel queuing problem where the decision-maker must choose to accept or reject randomly arriving tasks into a no buffer queue which are processed by N identical servers. Each task has a price, which is a positive real number, and a class. Each class of task has a different price distribution, service rate, and arrives according to an inhomogenous Poisson process. The objective is to decide which tasks to accept so that the total price of tasks processed is maximised over a finite horizon. We formulate the problem using a discrete time Markov Decision Process (MDP) with a hybrid state space. We show that the optimal value function has a specific structure, which enables us to solve the hybrid MDP exactly. Moreover, we rigorously prove that as the gap between successive decision epochs grows smaller, the discrete time solution approaches the optimal solution to the original continuous time problem. To improve the scalability of our approach to a greater number of servers and task classes, we present an approximation based on state abstraction. We validate our approach on synthetic data, as well as a real financial fraud data set, which is the motivating application for this work.

poster session 3 @ blue 4, poster session 12 @ blue 4, poster session 3, poster session 12

We study the problem of predicting the influence of a user in spreading fake (or real) news on social media. We propose a new model to address this problem which takes into account both user and tweet characteristics. We show that our model achieves an F1 score of 0.853, resp. 0.931, at predicting the influence of fake, resp. real, news spreaders, and outperforms existing baselines. We also investigate important features at predicting the influence of real vs. fake news spreaders.

poster session 5 @ blue 4, poster session 12 @ blue 4, poster session 5, poster session 12

For a monocular camera-based navigation system, if we could effectively explore scene geometric cues from RGB images, the geometry information will significantly facilitate the efficiency of the navigation system. Motivated by this, we propose a highly efficient point-goal navigation framework, dubbed Geo-Nav. In a nutshell, our Geo-Nav consists of two parts: a visual perception part and a navigation part. In the visual perception part, we firstly propose a Self-supervised Depth Estimation network (SDE) specially tailored for the monocular camera-based navigation agent. Our SDE learns a mapping from an RGB input image to its corresponding depth image by exploring scene geometric constraints in a self-consistency manner. Then, in order to achieve a representative visual representation from the RGB inputs and learned depth images, we propose a Cross-modality Pyramid Fusion module (CPF). Concretely, our CPF computes a patch-wise cross-modality correlation between different modal features and exploits the correlation to fuse and enhance features at each scale. Thanks to the patch-wise nature of our CPF, we can fuse feature maps at high resolution, allowing our visual network to perceive more image details. In the navigation part, our extracted visual representations are fed to a navigation policy network to learn how to map the visual representations to agent actions effectively. Extensive experiments on a widely-used multiple-room environment Gibson demonstrate that Geo-Nav outperforms the state-of-the-art in terms of efficiency and effectiveness.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

We address the overlooked unbiasedness in existing long-tailed classification methods: we find that their overall improvement is mostly attributed to the biased preference of "tail" over "head", as the test distribution is assumed to be balanced; however, when the test is as imbalanced as the long-tailed training data---let the test respect Zipf's law of nature---the "tail" bias is no longer beneficial overall because it hurts the "head" majorities. In this paper, we propose Cross-Domain Empirical Risk Minimization (xERM) for training an unbiased test-agnostic model to achieve strong performances on both test distributions, which empirically demonstrates that xERM fundamentally improves the classification by learning better feature representation rather than the "head vs. tail" game. Based on causality, we further theoretically explain why xERM achieves unbiasedness: the bias caused by the domain selection is removed by adjusting the empirical risks on the imbalanced domain and the balanced but unseen domain.

poster session 2 @ red 1, poster session 12 @ red 1, oral session 12 @ red 1, poster session 2, poster session 12, oral session 12

State-of-the-art deep NLP models have achieved impressive improvements on many tasks. However, they are found to be vulnerable to some perturbations. Before they are widely adopted, the fundamental issues of robustness need to be addressed. In this paper, we design a robustness enhancement method to defend against word substitution perturbation, whose basic idea is to fight perturbation with perturbation. We find that: although many well-trained deep models are not robust in the setting of the presence of adversarial samples, they satisfy weak robustness. That means they can handle most non-crafted perturbations well. Taking advantage of the weak robustness property of deep models, we utilize non-crafted perturbations to resist the adversarial perturbations crafted by attackers. Our method contains two main stages. The first stage is using randomized perturbation to conform the input to the data distribution. The second stage is using randomized perturbation to eliminate the instability of prediction results and enhance the robustness guarantee. Experimental results show that our method can significantly improve the ability of deep models to resist the state-of-the-art adversarial attacks while maintaining the prediction performance on the original clean data.

poster session 5 @ red 5, poster session 12 @ red 5, poster session 5, poster session 12

Current state-of-the-art deep learning based face recognition (FR) models require a large number of face identities for central training. However, due to the growing privacy awareness, it is prohibited to access the face images on user devices to continually improve face recognition models. Federated Learning (FL) is a technique to address the privacy issue, which can collaboratively optimize the model without sharing the data between clients. In this work, we propose a FL based framework called FedFR to improve the generic face representation in a privacy-aware manner. Besides, the framework jointly optimizes personalized models for the corresponding clients via the proposed Decoupled Feature Customization module. The client-specific personalized model can serve the need of optimized face recognition experience for registered identities at the local device. To the best of our knowledge, we are the first to explore the personalized face recognition in FL setup. The proposed framework is validated to be superior to previous approaches on several generic and personalized face recognition benchmarks with diverse FL scenarios. The source codes and our proposed personalized FR benchmark under FL setup are available at https://github.com/jackie840129/FedFR.

poster session 2 @ red 1, poster session 12 @ red 1, oral session 12 @ red 1, poster session 2, poster session 12, oral session 12

Although existing face anti-spoofing (FAS) methods achieve high accuracy in intra-domain experiments, their effects drop severely in cross-domain scenarios because of poor generalization. Recently, multifarious techniques have been explored, such as domain generalization and representation disentanglement. However, the improvement is still limited by two issues: 1) It is difficult to perfectly map all faces to a shared feature space. If faces from unknown domains are not mapped to the known region in the shared feature space, accidentally inaccurate predictions will be obtained. 2) It is hard to completely consider various spoof traces for disentanglement. In this paper, we propose a Feature Generation and Hypothesis Verification framework to alleviate the two issues. Above all, feature generation networks which generate hypotheses of real faces and known attacks are introduced for the first time in the FAS task. Subsequently, two hypothesis verification modules are applied to judge whether the input face comes from the real-face space and the real-face distribution respectively. Furthermore, some analyses of the relationship between our framework and Bayesian uncertainty estimation are given, which provides theoretical support for reliable defense in unknown domains. Experimental results show our framework achieves promising results and outperforms the state-of-the-art approaches on extensive public datasets.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

We propose an end-to-end image compression and analysis model with Transformers, targeting to the cloud-based image classification application. Instead of placing an existing Transformer-based image classification model directly after an image codec, we aim to redesign the Vision Transformer (ViT) model to perform image classification from the compressed features and facilitate image compression with the long-term information from the Transformer. Specifically, we first replace the patchify stem (i.e., image splitting and embedding) of the ViT model with a lightweight image encoder modelled by a convolutional neural network. The compressed features generated by the image encoder are injected convolutional inductive bias and are fed to the Transformer for image classification bypassing image reconstruction. Meanwhile, we propose a feature aggregation module to fuse the compressed features with the selected intermediate features of the Transformer, and feed the aggregated features to a deconvolutional neural network for image reconstruction. The aggregated features can obtain the long-term information from the self-attention mechanism of the Transformer and improve the compression performance. The rate-distortion-accuracy optimization problem is finally solved by a two-step training strategy. Experimental results demonstrate the effectiveness of the proposed model in both the image compression and the classification tasks.

poster session 5 @ red 2, poster session 12 @ red 2, oral session 5 @ red 2, poster session 5, poster session 12, oral session 5

Deep reinforcement learning (DRL) has demonstrated impressive performance in various gaming simulators and real-world applications. In practice, however, a DRL agent may receive faulty observation by abrupt interferences such as black-out, frozen-screen, and adversarial perturbation. How to design a resilient DRL algorithm against these rare but mission-critical and safety-crucial scenarios is an essential yet challenging task. In this paper, we consider a deep q-network (DQN) framework training with an auxiliary task of observational interferences such as artificial noises. Inspired by causal inference for observational interference, we propose a causal inference based DQN algorithm called causal inference Q-network (CIQ). We evaluate the performance of CIQ in several benchmark DQN environments with different types of interferences as auxiliary labels. Our experimental results show that the proposed CIQ method could achieve higher performance and more resilience against observational interferences.

poster session 6 @ blue 1, poster session 12 @ blue 1, poster session 6, poster session 12

Recent methods in speech and language technology pre-train very large models which are fine-tuned for specific tasks. However, the benefits of such large models are often limited to a few resource rich languages of the world. In this work, we make multiple contributions towards building ASR systems for low resource languages from the Indian subcontinent. First, we curate 16,000 hours of raw speech data for 40 Indian languages from a wide variety of domains including education, news, technology, and finance. Second, using this raw speech data we pre-train several variants of wav2vec style models for 40 Indian languages. Third, we analyze the pretrained models to find key features: code-book vectors of similar sounding phonemes are shared across languages, representations across layers are discriminative of the language family, and attention heads often pay attention within small local windows. Fourth, we fine-tune this model for downstream ASR for 9 languages and obtain state-of-the-art results on 3 public datasets, including on very low-resource languages such as Sinhala and Nepali. Our work establishes that multilingual pretraining is an effective strategy for building ASR systems for the linguistically diverse speakers of the Indian subcontinent.

poster session 5 @ red 5, poster session 12 @ red 5, oral session 12 @ red 5, poster session 5, poster session 12, oral session 12

Scheduling under uncertainty is an area of interest in artificial intelligence. We study the problem of Dynamic Controllability (DC) of Disjunctive Temporal Networks with Uncertainty (DTNU), which seeks a reactive scheduling strategy to satisfy temporal constraints in response to uncontrollable action durations. We introduce new semantics for reactive scheduling: Time-based Dynamic Controllability (TDC) and a restricted subset of TDC, R-TDC. We present a tree search approach to determine whether or not a DTNU is R-TDC. Moreover, we leverage the learning capability of a Graph Neural Network (GNN) as a heuristic for tree search guidance. Finally, we conduct experiments on a known benchmark on which we show R-TDC to retain significant completeness with regard to DC, while being faster to prove. This results in the tree search processing fifty percent more DTNU problems in R-TDC than the state-of-the-art DC solver does in DC with the same time budget. We also observe that GNN tree search guidance leads to substantial performance gains on benchmarks of more complex DTNUs, with up to eleven times more problems solved than the baseline tree search.

poster session 3 @ blue 4, poster session 12 @ blue 4, oral session 3 @ blue 4, poster session 3, poster session 12, oral session 3

Graph neural networks (GNNs) have been used to tackle the few-shot learning (FSL) problem and shown great potentials under the transductive setting. However under the inductive setting, existing GNN based methods are less competitive. This is because they use an instance GNN as a label propagation/classification module, which is jointly meta-learned with a feature embedding network. This design is problematic because the classifier needs to adapt quickly to new tasks while the embedding does not. To overcome this problem, in this paper we propose a novel hybrid GNN (HGNN) model consisting of two GNNs, an instance GNN and a prototype GNN. Instead of label propagation, they act as feature embedding adaptation modules for quick adaptation of the meta-learned feature embedding to new tasks. Importantly they are designed to deal with a fundamental yet often neglected challenge in FSL, that is, with only a handful of shots per class, any few-shot classifier would be sensitive to badly sampled shots which are either outliers or can cause inter-class distribution overlapping. Extensive experiments show that our HGNN obtains new state-of-the-art on three FSL benchmarks. The code and models are available at https://github.com/TianyuanYu/HGNN.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

Many deep learning tasks need to deal with graph data (e.g., social networks, protein structures, code ASTs). Due to the importance of these tasks, people turned to Graph Neural Networks (GNNs) as the de facto method for machine learning on graph data. GNNs have become widely applied due to their convincing performance. Unfortunately, one major barrier to using GNNs is that GNNs require substantial time and resources to train. Recently, a new method for learning on graph data is \textit{Graph Neural Tangent Kernel (GNTK)}~\cite{dhs+19}. GNTK is an application of Neural Tangent Kernel (NTK)~\cite{jgh18} (a kernel method) on graph data, and solving NTK regression is equivalent to using gradient descent to train an infinite-wide neural network. The key benefit of using GNTK is that, similar to any kernel method, GNTK's parameters can be solved directly in a single step, avoiding time-consuming gradient descent. Meanwhile, sketching has become increasingly used in speeding up various optimization problems, including solving kernel regression. Given a kernel matrix of $n$ graphs, using sketching in solving kernel regression can reduce the running time to $o(n^3)$. But unfortunately such methods usually requires extensive knowledge about the kernel matrix beforehand, while in the case of GNTK we find that the construction of the kernel matrix is already $O(n^2N^4)$, assuming each graph has $N$ nodes. The kernel matrix construction time can be a major performance bottleneck when the size of graphs $N$ increases. A natural question to ask is thus whether we can speed up the kernel matrix construction to improve GNTK regression's end-to-end running time. This paper provides the first algorithm to construct the kernel matrix in $o(n^2N^3)$ running time.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

Air pollution is one of the leading causes of mortality globally, resulting in millions of deaths each year. Efficient monitoring is important to measure exposure and enforce legal limits. New low-cost sensors can be deployed in greater numbers and in more varied locations, motivating the problem of efficient automated placement. Previous work suggests Bayesian optimisation is an appropriate method, but only considered a satellite data set, with data aggregated over all altitudes. It is ground-level pollution, that humans breathe, which matters most. We improve on those results using hierarchical models and evaluate our models on urban pollution data in London to show that Bayesian optimisation can be successfully applied to the problem.

poster session 3 @ red 6, poster session 12 @ red 6, poster session 3, poster session 12

The success of the state-of-the-art video deblurring methods stems mainly from implicit or explicit estimation of alignment among the adjacent frames for latent video restoration. However, due to the influence of the blur effect, estimating the alignment information from the blurry adjacent frames is not a trivial task. Inaccurate estimations will interfere the following frame restoration. Instead of estimating alignment information, we propose a simple and effective deep Recurrent Neural Network with Multi-scale Bi-directional Propagation (RNN-MBP) to effectively propagate and gather the information from unaligned neighboring frames for better video deblurring. Specifically, we build a Multi-scale Bi-directional Propagation~(MBP) module with two U-Net RNN cells which can directly exploit the inter-frame information from unaligned neighboring hidden states by integrating them in different scales. Moreover, to better evaluate the proposed algorithm and existing state-of-the-art methods on real-world blurry scenes, we also create a Real-World Blurry Video Dataset (RBVD) by a well-designed Digital Video Acquisition System (DVAS) and use it as the training and evaluation dataset. Extensive experimental results demonstrate that the proposed RBVD dataset effectively improve the performance of existing algorithms on real-world blurry videos, and the proposed algorithm performs favorably against the state-of-the-art methods on three typical benchmarks.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

Existing CNNs-Based RGB-D salient object detection (SOD) networks are all required to be pretrained on the ImageNet to learn the hierarchy features which helps provide a good initialization. However, the collection and annotation of large-scale datasets are time-consuming and expensive. In this paper, we utilize self-supervised representation learning (SSL) to design two pretext tasks: the cross-modal auto-encoder and the depth-contour estimation. Our pretext tasks require only a few and unlabeled RGB-D datasets to perform pretraining, which makes the network capture rich semantic contexts and reduce the gap between two modalities, thereby providing an effective initialization for the downstream task. In addition, for the inherent problem of cross-modal fusion in RGB-D SOD, we propose a consistency-difference aggregation (CDA) module that splits a single feature fusion into multi-path fusion to achieve an adequate perception of consistent and differential information. The CDA module is general and suitable for cross-modal and cross-level feature fusion. Extensive experiments on six benchmark datasets show that our self-supervised pretrained model performs favorably against most state-of-the-art methods pretrained on ImageNet.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

Multi-person pose estimation methods generally follow top-down and bottom-up paradigms, both of which can be considered as two-stage approaches thus leading to the high computation cost and low efficiency. Towards a compact and efficient pipeline for multi-person pose estimation task, in this paper, we propose to represent the human parts as points and present a novel body representation, which leverages an adaptive point set including the human center and seven human-part related points to represent the human instance in a more fine-grained manner. The novel representation is more capable of capturing the various pose deformation and adaptively factorizes the long-range center-to-joint displacement thus delivers a single-stage differentiable network to more precisely regress multi-person pose, termed as AdaptivePose. For inference, our proposed network eliminates the grouping as well as refinements and only needs a single-step disentangling process to form multi-person pose. Without any bells and whistles, we achieve the best speed-accuracy trade-offs of 67.4% AP / 29.4 fps with DLA-34 and 71.3% AP / 9.1 fps with HRNet-W48 on COCO test-dev dataset.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

Modern generative models achieve excellent quality in a variety of tasks including image or text generation and chemical molecule modeling. However, existing methods often lack the essential ability to generate examples with requested properties, such as the age of the person in the photo or the weight of the generated molecule. Incorporating such additional conditioning factors would require rebuilding the entire architecture and optimizing the parameters from scratch. Moreover, it is difficult to disentangle selected attributes so that to perform edits of only one attribute while leaving the others unchanged. To overcome these limitations we propose PluGeN (Plugin Generative Network), a simple yet effective generative technique that can be used as a plugin to pre-trained generative models. The idea behind our approach is to transform the entangled latent representation using a flow-based module into a multi-dimensional space where the values of each attribute are modeled as an independent one-dimensional distribution. In consequence, PluGeN can generate new samples with desired attributes as well as manipulate labeled attributes of existing examples. Due to the disentangling of the latent representation, we are even able to generate samples with rare or unseen combinations of attributes in the dataset, such as a young person with gray hair, men with make-up, or women with beards. We combined PluGeN with GAN and VAE models and applied it to conditional generation and manipulation of images and chemical molecule modeling. Experiments demonstrate that PluGeN preserves the quality of backbone models while adding the ability to control the values of labeled attributes.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

Visual scenes are extremely rich in diversity, not only because there are infinite combinations of objects and background, but also because the observations of the same scene may vary greatly with the change of viewpoints. When observing a visual scene that contains multiple objects from multiple viewpoints, humans are able to perceive the scene in a compositional way from each viewpoint, while achieving the so-called ``object constancy'' across different viewpoints, even though the exact viewpoints are untold. This ability is essential for humans to identify the same object while moving and to learn from vision efficiently. It is intriguing to design models that have the similar ability. In this paper, we consider a novel problem of learning compositional scene representations from multiple unspecified viewpoints without using any supervision, and propose a deep generative model which separates latent representations into a viewpoint-independent part and a viewpoint-dependent part to solve this problem. To infer latent representations, the information contained in different viewpoints is iteratively integrated by neural networks. Experiments on several specifically designed synthetic datasets have shown that the proposed method is able to effectively learn from multiple unspecified viewpoints.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

Off-the-shelf single-stage multi-person pose regression methods generally leverage the instance score (i.e., confidence of the instance localization) to indicate the pose quality for selecting the pose candidates. We consider that there are two gaps involved in existing paradigm: 1) The instance score is not well interrelated with the pose regression quality. 2) The instance feature representation, which is used for predicting the instance score, does not explicitly encode the structural pose information to predict the reasonable score that represents pose regression quality. To address the aforementioned issues, we propose to learn the pose regression quality-aware representation. Concretely, for the first gap, instead of using the previous instance confidence label (e.g., discrete {1,0} or Gaussian representation) to denote the position and confidence for person instance, we firstly introduce the Consistent Instance Representation (CIR) that unifies the pose regression quality score of instance and the confidence of background into a pixel-wise score map to calibrates the inconsistency between instance score and pose regression quality. To fill the second gap, we further present the Query Encoding Module (QEM) including the Keypoint Query Encoding (KQE) to encode the positional and semantic information for each keypoint and the Pose Query Encoding (PQE) which explicitly encodes the predicted structural pose information to better fit the Consistent Instance Representation (CIR). By using the proposed components, we significantly alleviate the above gaps. Our method outperforms previous single-stage regression-based even bottom-up methods and achieves the state-of-the-art result of 71.7 AP on MS COCO test-dev set.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

The goal of the crime forecasting problem is to predict different types of crimes for each geographical region (like a neighborhood or censor tract) in the near future. Since nearby regions usually have similar socioeconomic characteristics which indicate similar crime patterns, recent state-of-the-art solutions constructed a distance-based region graph and utilized Graph Neural Network (GNN) techniques for crime forecasting, because the GNN techniques could effectively exploit the latent relationships between neighboring region nodes in the graph if the edges reveal high dependency or correlation. However, this distance-based pre-defined graph can not fully capture crime correlation between regions that are far from each other but share similar crime patterns. Hence, to make a more accurate crime prediction, the main challenge is to learn a better graph that reveals the dependencies between regions in crime occurrences and meanwhile captures the temporal patterns from historical crime records. To address these challenges, we propose an end-to-end graph convolutional recurrent network called HAGEN with several novel designs for crime prediction. Specifically, our framework could jointly capture the crime correlation between regions and the temporal crime dynamics by combining an adaptive region graph learning module with the Diffusion Convolution Gated Recurrent Unit (DCGRU). Based on the homophily assumption of GNN (i.e., graph convolution works better where neighboring nodes share the same label), we propose a homophily-aware constraint to regularize the optimization of the region graph so that neighboring region nodes on the learned graph share similar crime patterns, thus fitting the mechanism of diffusion convolution. Empirical experiments and comprehensive analysis on two real-world datasets showcase the effectiveness of HAGEN.

poster session 6 @ blue 5, poster session 12 @ blue 5, oral session 6 @ blue 5, poster session 6, poster session 12, oral session 6

In high-stake scenarios like medical treatment and auto-piloting, it's risky or even infeasible to collect online experimental data to train the agent. Simulation-based training can alleviate this issue, but probably suffers from its inherent mismatches from the simulator and real environment. It is therefore imperative to utilize the simulator to learn a robust policy for the real-world deployment. In this work, we consider the policy learning for Robust Markov Decision Processes (RMDP), where the agent tries to seek a robust policy with respect to unexpected perturbations on the environments. Specifically, we focus on the setting where the training environment can be characterized as a generative model and a constrained perturbation can be added to the model during testing. Our goal is to identify a near-optimal robust policy for the perturbed testing environment, which introduces additional technical difficulties as we need to simultaneously estimate the training environment uncertainty from samples and find the worst-case perturbation for testing. To solve this issue, we propose a generic method which formalizes the perturbation as an opponent to obtain a two-player zero-sum game, and further show that the Nash Equilibrium corresponds to the robust policy. We prove that, with a polynomial number of samples from the generative model, our algorithm can find a near-optimal robust policy with a high probability. Our method is able to deal with general perturbations under some mild assumptions and can also be extended to more complex problems like robust partial-observable Markov decision process, thanks to the game-theoretical formulation.

poster session 6 @ blue 1, poster session 12 @ blue 1, poster session 6, poster session 12

Molecular representation learning contributes to multiple downstream tasks such as molecular property prediction and drug design. To properly represent molecules, graph contrastive learning is a promising paradigm as it utilizes self-supervision signals and has no requirements for human annotations. However, prior works fail to incorporate fundamental domain knowledge into graph semantics and thus ignore the correlations between atoms that have common attributes but are not directly connected by bonds. To address these issues, we construct a Chemical Element Knowledge Graph (KG) to summarize microscopic associations between elements and propose a novel Knowledge-enhanced Contrastive Learning (KCL) framework for molecular representation learning. KCL framework consists of three modules. The first module, knowledge-guided graph augmentation, augments the original molecular graph based on the Chemical Element KG. The second module, knowledge-aware graph representation, extracts molecular representations with a common graph encoder for the original molecular graph and a Knowledgeaware Message Passing Neural Network (KMPNN) to encode complex information in the augmented molecular graph. The final module is a contrastive objective, where we maximize agreement between these two views of molecular graphs. Extensive experiments demonstrated that KCL obtained superior performances against state-of-the-art baselines on eight molecular datasets. Visualization experiments properly interpret what KCL has learned from atoms and attributes in the augmented molecular graphs. Our codes and data are available in supplementary materials.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

Correlated Equilibrium is a solution concept that is more general than Nash Equilibrium (NE) and can lead to outcomes with better social welfare. However, its natural extension to the sequential setting, the Extensive Form Correlated Equilibrium (EFCE), requires a quadratic amount of space to solve, even in restricted settings without randomness in nature. To alleviate these concerns, we apply subgame resolving, a technique extremely successful in finding NE in zero-sum games to solving general-sum EFCEs. Subgame resolving refines a correlation plan in an online manner: instead of solving for the full game upfront, it only solves for strategies in subgames that are reached in actual play, resulting in significant computational gains. In this paper, we (i) lay out the foundations to quantify the quality of a refined strategy, in terms of the social welfare and exploitability of correlation plans, (ii) show that EFCEs possess a sufficient amount of independence between subgames to perform resolving efficiently, and (iii) provide two algorithms for resolving, one using linear programming and the other based on regret minimization. Both methods guarantee safety, i.e., they will never be counterproductive. Our methods are the first time an online method has been applied to the correlated, general-sum setting.

poster session 1 @ blue 6, poster session 12 @ blue 6, oral session 12 @ blue 6, poster session 1, poster session 12, oral session 12

Under data distributions which may be heavy-tailed, many stochastic gradient-based learning algorithms are driven by feedback queried at points with almost no performance guarantees on their own. Here we explore a modified ``anytime online-to-batch'' mechanism which for smooth objectives admits high-probability error bounds while requiring only lower-order moment bounds on the stochastic gradients. Using this conversion, we can derive a wide variety of ``anytime robust'' procedures, for which the task of performance analysis can be effectively reduced to regret control, meaning that existing regret bounds (for the bounded gradient case) can be robustified and leveraged in a straightforward manner. As a direct takeaway, we obtain an easily implemented stochastic gradient-based algorithm for which all queried points formally enjoy sub-Gaussian error bounds, and in practice show noteworthy gains on real-world data applications.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

Citing and describing related literature are crucial to scientific writing. Many existing approaches show encouraging performance in citation recommendation, but are unable to accomplish the more challenging and onerous task of citation text generation. In this paper, we propose a novel disentangled representation based model DisenCite to automatically generate the citation text through integrating paper text and citation graph. A key novelty of our method compared with existing approaches is to generate context-specific citation text, empowering the generation of different types of citations for the same paper. In particular, we first build and make available a graph enhanced contextual citation dataset (GCite) with 25K edges in different types characterized by citation contained sections over 4.8K research papers. Based on this dataset, we encode each paper according to both textual contexts and structure information in the heterogeneous citation graph. The resulted paper representations are then disentangled by the mutual information regularization between this paper and its neighbors in graph. Extensive experiments demonstrate the superior performance of our method comparing to state-of-the-art approaches. We further conduct ablation and case studies to reassure that the improvement of our method comes from generating the context-specific citation through incorporating the citation graph.

poster session 5 @ red 5, poster session 12 @ red 5, oral session 5 @ red 5, poster session 5, poster session 12, oral session 5

We consider the single image super-resolution (SISR) problem, where a high-resolution (HR) image is generated based on a low-resolution (LR) input. Recently, generative adversarial networks (GANs) become popular to hallucinate details. Most methods along this line rely on a predefined single-LR-single-HR mapping, which is not flexible enough for the ill-posed SISR task. Also, GAN-generated fake details may often undermine the realism of the whole image. We address these issues by proposing best-buddy GANs (Beby-GAN) for rich-detail SISR. Relaxing the rigid one-to-one constraint, we allow the estimated patches to dynamically seek trustworthy surrogates of supervision during training, which is beneficial to producing more reasonable details. Besides, we propose a region-aware adversarial learning strategy that directs our model to focus on generating details for textured areas adaptively. Extensive experiments justify the effectiveness of our method. An ultra-high-resolution 4K dataset is also constructed to facilitate future super-resolution research.

poster session 5 @ red 2, poster session 12 @ red 2, oral session 12 @ red 2, poster session 5, poster session 12, oral session 12

A recent report of Littmann [Commun. ACM '21] outlines the existence and the fatal impact of collusion rings in academic peer reviewing. We introduce and analyze the problem Cycle-Free Reviewing that aims at finding a review assignment without the following kind of collusion ring: A sequence of reviewers each reviewing a paper authored by the next reviewer in the sequence (with the last reviewer reviewing a paper of the first), thus creating a review cycle where each reviewer gives favorable reviews. As a result, all papers in that cycle have a high chance of acceptance independent of their respective scientific merit. We observe that review assignments computed using a standard Linear Programming approach typically admit many short review cycles. On the negative side, we show that Cycle-Free Reviewing is NP-hard in various restricted cases (i.e., when every author is qualified to review all papers and one wants to prevent that authors review each other's or their own papers or when every author has only one paper and is only qualified to review few papers). On the positive side, among others, we show that, in some realistic settings, an assignment without any review cycles of small length always exists. This result also gives rise to an efficient heuristic for computing (weighted) cycle-free review assignments, which we show to be of excellent quality in practice.

poster session 1 @ blue 6, poster session 12 @ blue 6, poster session 1, poster session 12

In hedonic games, a set of $n$ agents, having preferences over all possible coalition structures, needs to agree on a stable outcome. In this work, we initiate the study of hedonic games with fixed-size coalitions, where the set of possible coalition structures is restricted as follows: there are $k$ coalitions, each coalition has a fixed size, and the sum of the sizes of all coalitions equals $n$. We focus on the basic model of additively separable hedonic games with symmetric preferences, where an agent's preference is captured by a utility function which sums up a contribution due to any other agent in the same coalition. In this setting, an outcome is stable if no pair of agents can exchange coalitions and improve their utilities. Conditioned on the definition of improvement, three stability notions arise: {\em swap stability under transferable utilities}, which requires to improve the sum of the utilities of both agents, {\em swap stability}, which requires to improve the utility of one agent without decreasing the utility of the other one, and {\em strict swap stability}, requiring to improve the utilities of both agents simultaneously. We analyse the fundamental questions of existence, complexity and efficiency of stable outcomes, and that of complexity of a social optimum.

poster session 1 @ blue 6, poster session 12 @ blue 6, poster session 1, poster session 12

For users to trust planning algorithms, they must be able to understand the planner's outputs and the reasons for each action selection. This output does not tend to be user-friendly, often consisting of sequences of parametrised actions or task networks. And these may not be practical for non-expert users who may find it easier to read natural language descriptions. In this paper, we propose PlanVerb, a domain and planner-independent method for the verbalization of task plans. It is based on semantic tagging of actions and predicates. Our method can generate natural language descriptions of plans including causal explanations. The verbalized plans can be summarized by compressing the actions that act on the same parameters. We further extend the concept of verbalization space, previously applied to robot navigation, and apply it to planning to generate different kinds of plan descriptions for different user requirements. Our method can deal with PDDL and RDDL domains, provided that they are tagged accordingly. Our user survey evaluation shows that users can read our automatically generated plan descriptions and that the explanations help them answer questions about the plan.

poster session 3 @ blue 4, poster session 12 @ blue 4, poster session 3, poster session 12

Various deepfake detectors have been proposed, but challenges still exist to detect images of unknown categories or GAN models outside of the training settings. Such issues arise from the overfitting issue, which we discover from our own analysis and the previous studies to originate from the frequency-level artifacts in generated images. We find that ignoring the frequency-level artifacts can improve the detector's generalization across various GAN models, but it can reduce the model's performance for the trained GAN models. Thus, we design a framework to generalize the deepfake detector for both the known and unseen GAN models. Our framework generates the frequency-level perturbation maps to make the generated images indistinguishable from the real images. By updating the deepfake detector along with the training of the perturbation generator, our model is trained to detect the frequency-level artifacts at the initial iterations and consider the image-level irregularities at the last iterations. For experiments, we design new test scenarios varying from the training settings in GAN models, color manipulations, and object categories. Numerous experiments validate the state-of-the-art performance of our deepfake detector.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Reference-based super-resolution (RefSR) has made significant progress in producing realistic textures using an external reference (Ref) image. However, existing RefSR methods obtain high-quality correspondence matchings consuming quadratic computation resources with respect to the input size, limiting its application. Moreover, these approaches usually suffer from scale misalignments between the low-resolution (LR) image and Ref image. In this paper, we propose an Accelerated Multi-Scale Aggregation network (AMSA) for Reference-based Super-Resolution, including Coarse-to-Fine Embedded PatchMatch (CFE-PatchMatch) and Multi-Scale Dynamic Aggregation (MSDA) module. To improve matching efficiency, we design a novel Embedded PatchMacth scheme with random samples propagation, which involves end-to-end training with asymptotic linear computational cost to the input size. To further reduce computational cost and speed up convergence, we apply the coarse-to-fine strategy on Embedded PatchMacth constituting CFE-PatchMatch. To fully leverage reference information across multiple scales and enhance robustness to scale misalignment, we develop the MSDA module consisting of Dynamic Aggregation and Multi-Scale Aggregation. The Dynamic Aggregation corrects minor scale misalignment by dynamically aggregating features, and the Multi-Scale Aggregation brings robustness to large scale misalignment by fusing multi-scale information. Experimental results show that the proposed AMSA achieves superior performance over state-of-the-art approaches on both quantitative and qualitative evaluations.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

An activation function is a crucial component of a neural network that introduces non-linearity in the network. The state-of-the-art performance of a neural network depends also on the perfect choice of an activation function. We propose two novel non-monotonic smooth trainable activation functions, called ErfAct and Pserf. Experiments suggest that the proposed functions improve the network performance significantly compared to the widely used activations like ReLU, Swish, and Mish. Replacing ReLU by ErfAct and Pserf, we have 5.68% and 5.42% improvement for top-1 accuracy on Shufflenet V2 (2.0x) network in CIFAR100 dataset, 2.11% and 1.96% improvement for top-1 accuracy on Shufflenet V2 (2.0x) network in CIFAR10 dataset, 1.0%, and 1.0% improvement on mean average precision (mAP) on SSD300 model in Pascal VOC dataset.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

The Elo rating system is widely adopted to evaluate the skills of (chess) game and sports players. Recently it has been also integrated into machine learning algorithms in evaluating the performance of computerised AI agents. However, an accurate estimation of the Elo rating (for the top players) often requires many rounds of competitions, which can be expensive to carry out. In this paper, to minimize the number of comparisons and to improve the sample efficiency of the Elo evaluation (for top players), we propose an efficient online match scheduling algorithm. Specifically, we identify and match the top players through a dueling bandits framework and tailor the bandit algorithm to the gradient-based update of Elo. We show that it reduces the per-step memory and time complexity to constant, compared to the traditional likelihood maximization approaches requiring $O(t)$ time. Our algorithm has a regret guarantee of $\tilde{O}(\sqrt{T}$)\footnote{$\tilde{O}$ ignores poly-logarithmic factors}, sublinear in the number of competition rounds and has been extended to the multidimensional Elo ratings for handling intransitive games. We empirically demonstrate that our method achieves superior convergence speed and time efficiency on a variety of gaming tasks.

poster session 6 @ blue 1, poster session 12 @ blue 1, poster session 6, poster session 12

Question answering over semi-structured tables has attracted significant attention in the NLP community. However, most of the existing work focus on questions that can be answered with short-form answer, i.e. the answer is often a table cell or aggregation of multiple cells. This can mismatch with the intents of users who want to ask more complex questions that require free-form answers such as explanations. To bridge the gap, most recently, pre-trained sequence-to-sequence language models such as T5 are used for generating free-form answers based on the question and table inputs. However, these pre-trained language models have weaker encoding abilities over table cells and schema. To mitigate this issue, in this work, we present an intermediate pre-training framework, Generation-focused Table-based Intermediate Pre-training (GENTAP), that jointly learns representations of natural language questions and tables. GENTAP learns to generate via two training objectives to enhance the question understanding and table representation abilities for complex questions. Based on experimental results, models that leverage GENTAP framework outperform the existing baselines on FETAQA benchmark. The pre-trained models are not only useful for free-form question answering, but also for few-shot data-to-text generation task, thus showing good transfer ability by obtaining new state-of-the-art results.

poster session 3 @ red 4, poster session 12 @ red 4, oral session 12 @ red 4, poster session 3, poster session 12, oral session 12

We investigate the computational complexity of mining guarded clauses from clausal datasets through the framework of inductive logic programming (ILP). We show that learning guarded clauses is NP-complete and thus one step below the Sigma2-complete task of learning Horn clauses on the polynomial hierarchy. Motivated by practical applications on large datasets we identify a natural tractable fragment of the problem. Finally, we also generalise all of our results to k-guarded clauses for constant k.

poster session 6 @ blue 5, poster session 12 @ blue 5, oral session 12 @ blue 5, poster session 6, poster session 12, oral session 12

Cooperative Pickup and Delivery Problem (PDP), as a variant of the typical Vehicle Routing Problems (VRP), is an important formulation in many real-world applications, such as on-demand delivery, industrial warehousing, etc. It is of great importance to efficiently provide high-quality solutions of cooperative PDP. However, it is not trivial to provide effective solutions directly due to two major challenges: 1) the structural dependency between pickup and delivery pairs require explicit modeling and representation. 2) the cooperation between different vehicles is highly related to the solution exploration and difficult to model. In this paper, we propose a novel multi-agent reinforcement learning based framework to solve the cooperative PDP (MAPDP). First, we design a paired context embedding to well measure the dependency of different nodes considering their structural limits. Second, we utilize cooperative multi-agent decoders to leverage the decision dependence among different vehicle agents based on a special communication embedding. Third, we design a novel cooperative A2C algorithm to train the integrated model. We conduct extensive experiments on a randomly generated dataset and a real-world dataset. Experiments result shown that the proposed MAPDP outperform all other baselines by at least 1.64\% in all settings, and shows significant computation speed during solution inference.

poster session 3 @ blue 4, poster session 12 @ blue 4, poster session 3, poster session 12

This paper considers a multi-state Log Gaussian Cox Process (`"LGCP'') on a graph, where transmissions amongst states are calibrated using a non-parametric approach. We thus consider multi-output LGCPs and introduce numerical approximations to compute posterior distributions extremely quickly and in a completely transparent and reproducible fashion. The model is tested on historical data and shows very good performance.

poster session 6 @ blue 1, poster session 12 @ blue 1, poster session 6, poster session 12

We consider a general task called partial Wasserstein covering with the goal of providing information on what patterns are not being taken into account in a dataset (e.g., dataset used during development) compared to another (e.g., dataset obtained from actual applications). We model this task as a discrete optimization problem with partial Wasserstein divergence as an objective function. Although this problem is NP-hard, we prove that it satisfies the submodular property, allowing us to use a greedy algorithm with a 0.63 approximation. However, the greedy algorithm is still inefficient because it requires solving linear programming for each objective function evaluation. To overcome this inefficiency, we propose quasi-greedy algorithms, which consist of a series of techniques for acceleration such as sensitivity analysis based on strong duality and the so-called $C$-transform in the optimal transport field. Experimentally, we demonstrate that we can efficiently fill in the gaps between the two datasets, and find missing scene in real driving scene datasets.

poster session 6 @ blue 1, poster session 12 @ blue 1, poster session 6, poster session 12

Knowledge of the changing traffic is critical in risk management. Customs offices worldwide have traditionally relied on local resources to accumulate such knowledge and detect tax frauds. This naturally poses countries with weak infrastructure to become tax havens of potentially illicit trades. The current paper proposes DAS, a memory bank platform to facilitate knowledge sharing across multi-national customs administrations to support each other. We propose a domain adaptation method to share transferable knowledge of frauds as prototypes while safeguarding the local trade information. Data encompassing over 8 million import declarations have been used to test the feasibility of this new system, which shows that participating countries may benefit up to 2-11 times in fraud detection with the help of shared knowledge. We discuss implications for substantial tax revenue potential and strengthened policy against illicit trades.

poster session 5 @ red 6, poster session 12 @ red 6, poster session 5, poster session 12

We study fair and efficient allocation of divisible goods, in an online manner, among n agents. The goods arrive online in a sequence of T time periods. The agents' values for a good are revealed only after its arrival, and the online algorithm needs to fractionally allocate the good, immediately and irrevocably, among the agents. Towards a unifying treatment of fairness and economic efficiency objectives, we develop an algorithmic framework for finding online allocations to maximize the generalized mean of the values received by the agents. In particular, working with the assumption that each agent's value for the grand bundle of goods is appropriately scaled, we address online maximization of p-mean welfare. Parameterized by an exponent term p in (-infinity, 1], these means encapsulate a range of welfare functions, including social welfare (p=1), egalitarian welfare (p to -infinity), and Nash social welfare (p to 0). We present a simple algorithmic template that takes a threshold as input and, with judicious choices for this threshold, leads to both universal and tailored competitive guarantees. First, we show that one can compute online a single allocation that O (n^{1/2} log n)-approximates the optimal p-mean welfare for all p<=1. The existence of such a universal allocation is interesting in and of itself. Moreover, this universal guarantee achieves essentially tight competitive ratios for specific values of p. Next, we obtain improved competitive ratios for different ranges of p by executing our algorithm with p-specific thresholds, e.g., we provide O(log^3 n)-competitive ratio for all p in (-1/log 2n, 1). We complement our positive results by establishing lower bounds to show that our guarantees are essentially tight for a wide range of the exponent parameter.

poster session 1 @ blue 6, poster session 12 @ blue 6, oral session 1 @ blue 6, poster session 1, poster session 12, oral session 1

Dynamic Vision Sensor (DVS) can asynchronously output the events reflecting apparent motion of objects with microsecond resolution, and shows great application potential in monitoring and other fields. However, the output event stream of existing DVS inevitably contains background activity noise (BA noise) due to dark current and junction leakage current, which will affect the temporal correlation of objects, resulting in deteriorated motion estimation performance. Particularly, the existing filter-based denoising methods cannot be directly applied to suppress the noise in event stream, since there is no spatial correlation. To address this issue, this paper presents a novel progressive framework, in which a Motion Estimation (ME) module and an Event Denoising (ED) module are jointly optimized in a mutually reinforced manner. Specifically, based on the maximum sharpness criterion, ME module divides the input event into several segments by adaptive clustering in a motion compensating warp field, and captures the temporal correlation of event stream according to the clustered motion parameters. Taking temporal correlation as guidance, ED module calculates the confidence that each event belongs to real activity events, and transmits it to ME module to update energy function of motion segmentation for noise suppression. The two steps are iteratively updated until stable motion segmentation results are obtained. Extensive experimental results on both synthetic and real datasets demonstrate the superiority of our proposed approaches against the State-Of-The-Art (SOTA) methods.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

Text content created by humans or language models is often stolen or misused by adversaries. Tracing text provenance can help claim the ownership of text content or identify the malicious users who distribute misleading content like machine-generated fake news. There have been some attempts to achieve this, mainly based on watermarking techniques. Specifically, traditional text watermarking methods embed watermarks by slightly altering text format like line spacing and font, which, however, are fragile to cross-media transmissions like OCR. Considering this, natural language watermarking methods represent watermarks by replacing words in original sentences with synonyms from handcrafted lexical resources (e.g., WordNet), but they do not consider the substitution’s impact on the overall sentence's meaning. Recently, a transformer-based network was proposed to embed watermarks by modifying the unobtrusive words (e.g., function words), which also impair the sentence's logical and semantic coherence. Besides, one well-trained network fails on other different types of text content. To address the limitations mentioned above, we propose a natural language watermarking scheme based on context-aware lexical substitution (LS). Specifically, we employ BERT to suggest LS candidates by inferring the semantic relatedness between the candidates and the original sentence. Based on this, a selection strategy in terms of synchronicity and substitutability is further designed to test whether a word is exactly suitable for carrying the watermark signal. Extensive experiments demonstrate that, under both objective and subjective metrics, our watermarking scheme can well preserve the semantic integrity of original sentences and has a better transferability than existing methods. Besides, the proposed LS approach outperforms the state-of-the-art approach on the Stanford Word Substitution Benchmark.

poster session 5 @ red 5, poster session 12 @ red 5, oral session 12 @ red 5, poster session 5, poster session 12, oral session 12

Recognizing speech from silent lip movement, which is called lip reading, is a challenging task due to 1) the inherent information insufficiency of lip movement to fully represent the speech, and 2) the existence of homophenes that have similar lip movement with different pronunciations. In this paper, we try to alleviate the aforementioned two challenges in lip reading by proposing a Multi-head Visual-audio Memory (MVM). Firstly, MVM is trained with audio-visual datasets and remembers audio representations by modelling the inter-relationships of paired audio-visual representations. At the inference stage, visual input alone can extract the saved audio representation from the memory by examining the learned inter-relationships. Therefore, the lip reading model can complement the insufficient visual information with the extracted audio representations. Secondly, MVM is composed of multi-head key memories for saving visual features and one value memory for saving audio knowledge, which is designed to distinguish the homophenes. With the multi-head key memories, MVM extracts possible candidate audio features from the memory, which allows the lip reading model to consider the possibility of which pronunciations can be represented from the input lip movement. This also can be viewed as an explicit implementation of the one-to-many mapping of viseme-to-phoneme. Moreover, MVM is employed in multi-temporal levels to consider the context when retrieving the memory and distinguish the homophenes. Extensive experimental results verify the effectiveness of the proposed method in lip reading and in distinguishing the homophenes.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

Understanding events entails recognizing the structural and temporal orders between event mentions to build event structures/graphs for input documents. To achieve this goal, our work addresses the problems of subevent relation extraction (SRE) and temporal event relation extraction (TRE) that aim to predict subevent and temporal relations between two given event mentions/triggers in texts. Recent state-of-the-art methods for such problems have employed transformer-based language models (e.g., BERT) to induce effective contextual representations for input event mention pairs. However, a major limitation of existing transformer-based models for SRE and TRE is that they can only encode input texts of limited length (i.e., up to 512 sub-tokens in BERT), thus unable to effectively capture important context sentences that are farther away in the documents. In this work, we introduce a novel method to better model document-level context with important context sentences for event-event relation extraction. Our method seeks to identify the most important context sentences for a given entity mention pair in a document and pack them into shorter documents to be consume entirely by transformer-based language models for representation learning. The REINFORCE algorithm is employed to train models where novel reward functions are presented to capture model performance, and context-based and knowledge-based similarity between sentences for our problem. Extensive experiments demonstrate the effectiveness of the proposed method with state-of-the-art performance on benchmark datasets.

poster session 3 @ red 4, poster session 12 @ red 4, poster session 3, poster session 12

We consider datasets consisting of arbitrarily structured entities (e.g., molecules, sequences, graphs, etc) whose similarity can be assessed with a reproducing ker- nel (or a family thereof). These entities are assumed to additionally have a set of named attributes (e.g.: number_of_atoms, stock_price, etc). These attributes can be used to classify the structured entities in discrete sets (e.g., ‘number_of_atoms < 3’, ‘stock_price ≤ 100’, etc) and can effectively serve as Boolean predicates. Our goal is to use this side-information to provide explain- able kernel-based clustering. To this end, we propose a method which is able to find among all possible entity subsets that can be described as a conjunction of the available predicates either a) the optimal cluster within the Reproducing Kernel Hilbert Space, or b) the most anomalous subset within the same space. Our method works employs combinatorial optimisation via an adaptation of the Maximum-Mean-Discrepancy measure that captures the above intuition. Finally, we propose a criterion to select the optimal one out of a family of kernels in a way that preserves the available side-information. We provide several real world datasets that demonstrate the usefulness of our proposed method.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

Common image-text joint understanding techniques presume that image and associated text can universally be characterized by a single implicit model. However, co-occurring images and text can be related in qualitatively different ways, and explicitly modeling this could improve the performance of current joint understanding models. In this paper, we train a \cmcm for text-to-image retrieval task. Our analysis shows that models trained with image--text coherence relations can retrieve images originally paired with target text more often than coherence-agnostic models. We also show via human evaluation that the images retrieved by the proposed coherence-aware model are preferred over a coherence-agnostic baseline by a huge margin. Our findings provide insights into the ways that different modalities communicate and the role of coherence relations in capturing commonsense inferences in text and imagery.

poster session 5 @ red 5, poster session 12 @ red 5, poster session 5, poster session 12

Given an initial resource allocation, where some agents may envy others or where a different distribution of resources might lead to higher social welfare, our goal is to improve the allocation without reassigning resources. We consider a sharing concept allowing resources being shared with social network neighbors of the resource owners. To this end, we introduce a formal model that allows a central authority to compute an optimal sharing between neighbors based on an initial allocation. Advocating this point of view, we focus on the most basic scenario where a resource may be shared by two neighbors in a social network and each agent can participate in a bounded number of sharings. We present algorithms for optimizing utilitarian and egalitarian social welfare of allocations and for reducing the number of envious agents. In particular, we examine the computational complexity with respect to several natural parameters. Furthermore, we study cases with restricted social network structures and, among others, devise polynomial-time algorithms in path- and tree-like (hierarchical) social networks.

poster session 1 @ blue 6, poster session 12 @ blue 6, oral session 1 @ blue 6, poster session 1, poster session 12, oral session 1

Scene Graph Generation (SGG) aims to build a structured representation of a scene using objects and pairwise relationships, which benefits downstream tasks. However, current SGG methods usually suffer from sub-optimal scene graph generation because of the long-tailed distribution of training data. To address this problem, we propose Resistance Training using Prior Bias (RTPB) for the scene graph generation. Specifically, RTPB uses a distributed-based prior bias to improve models' detecting ability on less frequent relationships during training, thus improving the model generalizability on tail categories. In addition, to further explore the contextual information of objects and relationships, we design a contextual encoding backbone network, termed as Dual Transformer (DTrans). We perform extensive experiments on a very popular benchmark, VG150, to demonstrate the effectiveness of our method for the unbiased scene graph generation. In specific, our RTPB achieves an improvement of over 10\% under the mean recall when applied to current SGG methods. Furthermore, DTrans with RTPB outperforms nearly all state-of-the-art methods with a large margin.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

Bilinear pooling has achieved excellent performance in fine-grained recognition tasks. Nevertheless, high-dimensional bilinear features suffer from over-fitting and inefficiency. To alleviate these issues, compact bilinear pooling (CBP) methods were developed to generate low-dimensional features. Although the low-dimensional features from existing CBP methods enable high efficiency in subsequent classification, CBP methods themselves are inefficient. Thus, the inefficiency issue of the bilinear pooling is still unsolved. In this work, we propose an efficient compact bilinear pooling method to solve the inefficiency problem inherited in bilinear pooling thoroughly. It decomposes the huge-scale projection matrix into a two-level Kronecker product of several small-scale matrices. By exploiting the ``vec trick'' and the tensor modal product, we can obtain the compact bilinear feature through the decomposed projection matrices in a speedy manner. Systematic experiments on four public benchmarks using two backbones demonstrate the efficiency and effectiveness of the proposed method in fine-grained recognition.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

Translation-based approaches to planning allow for solving problems in complex and expressive formalisms via the means of highly efficient solvers for simpler formalisms. To be effective, these translations have to be constructed appropriately. The currently existing translation of the highly expressive formalism of HTN planning into the simplistic formalism of classical planning is not on par with the performance of current dedicated HTN planners. With our contributions in this paper, we close this gap: we describe new versions of the translation that reach the performance of state-of-the-art dedicated HTN planners. We present new translation techniques both for the special case of totally-ordered HTNs as well as for the general partially-ordered case. In the latter, we show hat our new translation is exponentially smaller than the previous one.

poster session 3 @ blue 4, poster session 12 @ blue 4, poster session 3, poster session 12

Exploring unknown environments is a fundamental task in many domains, e.g., robot navigation, network security, and internet search. We initiate the study of a learning-augmented variant of the classical, notoriously hard online graph exploration problem by adding access to machine-learned predictions. We propose an algorithm that naturally integrates predictions into the well-known Nearest Neighbor (NN) algorithm and significantly outperforms any known online algorithm if the prediction is of high accuracy while maintaining good guarantees when the prediction is of poor quality. We provide theoretical worst-case bounds that gracefully degrade with the prediction error, and we complement them by computational experiments that confirm our results. Further, we extend our concept to a general framework to robustify algorithms. By interpolating carefully between a given algorithm and NN, we prove new performance bounds that leverage the individual good performance on particular inputs while establishing robustness to arbitrary inputs.

poster session 3 @ blue 4, poster session 12 @ blue 4, poster session 3, poster session 12

More transformer blocks with residual connections have recently achieved impressive results on various tasks. To achieve better performance with fewer trainable parameters, recent methods are proposed to go shallower by parameter sharing or model compressing along with the depth. However, weak modeling capacity limits their performance. Contrastively, going wider by inducing more trainable matrixes and parameters would produce a huge model requiring advanced parallelism to train and inference. In this paper, we propose a parameter-efficient framework, going wider instead of deeper. Specially, following existing works, we adapt parameter sharing to compress along depth. But, such deployment would limit the performance. To maximize modeling capacity, we scale along model width by replacing feed-forward network (FFN) with mixture-of-experts (MoE). Across transformer blocks, instead of sharing normalization layers, we propose to use individual layernorms to transform various semantic representations in a more parameter-efficient way. To evaluate our plug-and-run framework, we design WideNet and conduct comprehensive experiments on popular computer vision and natural language processing benchmarks. On ImageNet-1K, our best model outperforms Vision Transformer (ViT) by $1.5\%$ with $0.72 \times$ trainable parameters. Using $0.46 \times$ and $0.13 \times$ parameters, our WideNet can still surpass ViT and ViT-MoE by $0.8\%$ and $2.1\%$, respectively. On four natural language processing datasets, WideNet outperforms ALBERT by $1.8\%$ on average and surpass BERT using factorized embedding parameterization by $0.8\%$ with fewer parameters.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

Understanding how a city’s physical appearance and environmental surroundings impact society traits, such as safety, is an essential issue in social artificial intelligence. To demonstrate the relationship, most existing studies utilize subjective human perceptual attributes, categorization only for a few violent crimes, and images taken from still shot images. These lead to difficulty in identifying location-specific characteristics for urban safety. In this work, to address this problem, we propose a large-scale dataset and a novel method by adopting a concept of “Deviance" which explains behaviors violating social norms, both formally (e.g. crime) and informally (e.g. civil complaints). We first collect a geo-tagged dataset consisting of incident report data for seven metropolitan cities, with corresponding sequential images around incident sites obtained from Google street view. We also design a convolutional neural network that learns spatio-temporal visual attributes of deviant streets. Experimental results show that our framework can reliably recognize real-world deviance in various cities. Furthermore, we analyze which visual attribute is important for deviance identification and severity estimation. We have released our dataset and source codes at our project page: https://deviance-project.github.io/DevianceNet/

poster session 3 @ red 6, poster session 12 @ red 6, poster session 3, poster session 12

The paper studies strategic abilities that rise from restrictions on the information sharing in multi-agent systems. The main technical result is a sound and complete logical system that describes the interplay between the knowledge and the strategic ability modalities.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

From ecology to atmospheric sciences, many academic disciplines deal with data characterized by intricate spatio-temporal complexities, the modeling of which often requires specialized approaches. Generative models of these data are of particular interest, as they enable a range of impactful downstream applications like simulation or creating synthetic training data. Recently, COT-GAN, a new GAN algorithm inspired by the theory of causal optimal transport (COT), was proposed in an attempt to improve generation of sequential data. However, the task of learning complex patterns over time and space requires additional knowledge of the specific data structures. In this study, we propose a novel loss objective combined with COT-GAN based on an autoregressive embedding to reinforce the learning of spatio-temporal dynamics. We devise SPATE (spatio-temporal association), a new metric measuring spatio-temporal autocorrelation. We compute SPATE for real and synthetic data samples and use it to compute an embedding loss that considers space-time interactions, nudging the GAN to learn outputs that are faithful to the observed dynamics. We test our new SPATE-GAN on a diverse set of spatio-temporal patterns: turbulent flows, log-Gaussian Cox processes and global weather data. We show that our novel embedding loss improves performance without any changes to the architecture of the GAN backbone, highlighting our model's increased capacity for capturing autoregressive structures.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

Controllers for autonomous systems that operate in safety-critical settings must account for stochastic disturbances. Such disturbances are often modeled as process noise, and common assumptions are that the underlying distributions are known and/or Gaussian. In practice, however, these assumptions may be unrealistic and can lead to poor approximations of the true noise distribution. We present a novel planning method that does not rely on any explicit representation of the noise distributions. In particular, we address the problem of computing a controller that provides probabilistic guarantees on safely reaching a target. First, we abstract the continuous system into a discrete-state model that captures noise by probabilistic transitions between states. As a key contribution, we adapt tools from the scenario approach to compute probably approximately correct (PAC) bounds on these transition probabilities, based on a finite number of samples of the noise. We capture these bounds in the transition probability intervals of a so-called interval Markov decision process (iMDP). This iMDP is robust against uncertainty in the transition probabilities, and the tightness of the probability intervals can be controlled through the number of samples. We use state-of-the-art verification techniques to provide guarantees on the iMDP, and compute a controller for which these guarantees carry over to the autonomous system. Realistic benchmarks show the practical applicability of our method, even when the iMDP has millions of states or transitions.

poster session 3 @ blue 4, poster session 12 @ blue 4, oral session 3 @ blue 4, poster session 3, poster session 12, oral session 3

We treat the problem of risk-aware control for stochastic shortest path (SSP) on Markov decision processes (MDP). Typically, expectation is considered for SSP, which however is oblivious to the incurred risk. We present an alternative view, instead optimizing conditional value-at-risk (CVaR), an established risk measure. We treat both Markov chains as well as MDP and introduce, through novel insights, two algorithms, based on linear programming and value iteration, respectively. Both algorithms offer precise and provably correct solutions. Evaluation of our prototype implementation shows that risk-aware control is feasible on several moderately sized models.

poster session 3 @ blue 4, poster session 12 @ blue 4, poster session 3, poster session 12

Amongst a variety of approaches aimed at making the learning procedure of neural networks more effective, the scientific community developed strategies to order the examples according to their estimated complexity, to distil knowledge from larger networks, or to exploit the principles behind adversarial machine learning. A different idea has been recently proposed, named Friendly Training, which consists in altering the input data by adding an automatically estimated perturbation, with the goal of facilitating the learning process of a neural classifier. The transformation progressively fades-out as long as training proceeds, until it completely vanishes. In this work we revisit and extend this idea, introducing a radically different and novel approach inspired by the effectiveness of neural generators in the context of Adversarial Machine Learning. We propose an auxiliary multi-layer network that is responsible of altering the input data to make them easier to be handled by the classifier at the current stage of the training procedure. The auxiliary network is trained jointly with the neural classifier, thus intrinsically increasing the 'depth' of the classifier, and it is expected to spot general regularities in the data alteration process. The effect of the auxiliary network is progressively reduced up to the end of training, when it is fully dropped and the classifier is deployed for applications. We refer to this approach as Neural Friendly Training. An extended experimental procedure involving several datasets and different neural architectures shows that Neural Friendly Training overcomes the originally proposed Friendly Training technique, improving the generalization of the classifier, especially in the case of noisy data.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

Multilingual neural machine translation architectures mainly differ in the number of sharing modules and parameters applied among languages. In this paper, and from an algorithmic perspective, we explore whether the chosen architecture, when trained with the same data, influences the level of gender bias. Experiments conducted in three language pairs show that language-specific encoder-decoders exhibit less bias than the shared architecture. We propose two methods for interpreting and studying gender bias in machine translation based on source embeddings and attention. Our analysis shows that, in the language-specific case, the embeddings encode more gender information, and their attention is more diverted. Both behaviors help in mitigating gender bias.

poster session 3 @ red 6, poster session 12 @ red 6, poster session 3, poster session 12

Detecting depression based on non-verbal behaviors has received great attention. However, most prior work on detecting depression mainly focused on detecting depressed individuals in laboratory settings, which are difficult to be generalized in practice. In addition, little attention has been paid to analyzing the non-verbal behaviors of depressed individuals in the wild. Therefore, in this paper, we present a multimodal depression dataset, D-Vlog, which consists of 961 vlogs (i.e., around 160 hours) collected from YouTube, which can be utilized in developing depression detection models based on the non-verbal behavior of individuals in real-world scenario. We develop a multimodal deep learning model that uses acoustic and visual features extracted from collected data to detect depression. Our proposed model employs the cross-attention mechanism to effectively capture the relationship across acoustic and visual features, and generates useful multimodal representations for depression detection. The extensive experimental results demonstrate that the proposed model significantly outperforms other baseline models. We believe our dataset and the proposed model are useful for analyzing and detecting depressed individuals based on non-verbal behavior.

poster session 5 @ red 6, poster session 12 @ red 6, poster session 5, poster session 12

InstaHide is a state-of-the-art mechanism for protecting private training images, which works by mixing multiple private images and modifying them such that their visual features are indistinguishable to the naked eye. In recent work, however, Carlini et al. show that it is possible to reconstruct private images from the encrypted dataset generated by InstaHide. Nevertheless, we demonstrate that Carlini et al.'s attack can be easily defeated by incorporating data augmentation into InstaHide. This leads to a natural question: is InstaHide with data augmentation secure? In this paper, we provide a negative answer to this question, by presenting an attack for recovering private images from the outputs of InstaHide even when data augmentation is present. The basic idea of our attack is to use a comparative network to identify encrypted images that are likely to correspond to the same private image, and then employ a fusion-denoising network for restoring the private image from the encrypted ones, taking into account the effects of data augmentation. Extensive experiments demonstrate the effectiveness of the proposed attack in comparison to Carlini et al.'s attack.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Recent studies indicate that deep neural networks (DNNs) are prone to show discrimination towards certain demographic groups. We observe that algorithmic discrimination can be explained by the high reliance of the models on fairness sensitive features. Motivated by this observation, we propose to achieve fairness by suppressing the DNN models from capturing the spurious correlation between those fairness sensitive features with the underlying task. Specifically, we firstly train a bias-only teacher model which is explicitly encouraged to maximally employ fairness sensitive features for prediction. The teacher model then counter-teaches a debiased student model so that the interpretation of the student model is orthogonal to the interpretation of the teacher model. The key idea is that since the teacher model relies explicitly on fairness sensitive features for prediction, the orthogonal interpretation loss enforces the student network to reduce its reliance on sensitive features and instead capture more task relevant features for prediction. Experimental analysis indicates that our framework substantially reduces the model's attention on fairness sensitive features. Experimental results on four datasets further validate that our framework has consistently improved the fairness with respect to three group fairness metrics, with a comparable or even better accuracy.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

One-class classification (OCC) aims to learn an effective data description to enclose all normal training samples and detect anomalies based on the deviation from the data description. Current state-of-the-art OCC models learn a compact normality description by hyper-sphere minimisation, but they often suffer from overfitting the training data, especially when the training set is small or contaminated with anomalous samples. To address this issue, we introduce the interpolated Gaussian descriptor (IGD) method, a novel OCC model that learns a one-class Gaussian anomaly classifier trained with adversarially interpolated training samples. The Gaussian anomaly classifier differentiates the training samples based on their distance to the Gaussian centre and the standard deviation of these distances, offering the model a discriminability w.r.t. the given samples during training. The adversarial interpolation is enforced to consistently learn a smooth Gaussian descriptor, even when the training data is small or contaminated with anomalous samples. This enables our model to learn the data description based on the representative normal samples rather than fringe or anomalous samples, resulting in significantly improved normality description. In extensive experiments on diverse popular benchmarks, including MNIST, Fashion MNIST, CIFAR10, MVTec AD and two medical datasets, IGD achieves better detection accuracy than current state-of-the-art models. IGD also shows better robustness in problems with small or contaminated training sets.

poster session 6 @ blue 2, poster session 12 @ blue 2, oral session 12 @ blue 2, poster session 6, poster session 12, oral session 12

Non-Local Attention (NLA) brings significant improvement for Single Image Super-Resolution (SISR) by leveraging intrinsic feature correlation in natural images. However, NLA gives noisy information large weights and consumes quadratic computation resources with respect to the input size, limiting its performance and application. In this paper, we propose a novel Efficient Non-Local Contrastive Attention (ENLCA) to perform long-range visual modeling and leverage more relevant non-local features. Specifically, ENLCA consists of two parts, Efficient Non-Local Attention (ENLA) and Sparse Aggregation. ENLA adopts the kernel method to approximate exponential function and obtains linear computation complexity. For Sparse Aggregation, we multiply inputs by an amplification factor to focus on informative features, yet the variance of approximation increases exponentially. Therefore, contrastive learning is applied to further separate relevant and irrelevant features. To demonstrate the effectiveness of ENLCA, we build an architecture called Efficient Non-Local Contrastive Network (ENLCN) by adding a few of our modules in a simple backbone. Extensive experimental results show that ENLCN reaches superior performance over state-of-the-art approaches on both quantitative and qualitative evaluations.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

The recently advanced unsupervised learning approaches use the siamese-like framework to compare two "views" from the same image for learning representations. Making the two views distinctive is a core to guarantee that unsupervised methods can learn meaningful information. However, such frameworks are sometimes fragile on overfitting if the augmentations used for generating two views are not strong enough, causing the over-confident issue on the training data. This drawback hinders the model from learning subtle variance and fine-grained information. To address this, in this work we aim to involve the distance concept on label space in the unsupervised learning and let the model be aware of the degree of similarity between positive or negative pairs through mixing the input space, to further work collaboratively for the input and loss spaces. Despite its conceptual simplicity, we show empirically that with the solution -- Unsupervised image mixtures (Un-Mix), we can learn more robust and generalized representations from the transformed input and corresponding new label space. Extensive experiments are conducted on CIFAR-10, CIFAR-100, STL-10, Tiny ImageNet and standard ImageNet with popular unsupervised methods SimCLR, BYOL, MoCo V1&V2, SwAV, etc. Our proposed image mixture and label assignment strategy can obtain consistent improvement by 1~3% following exactly the same hyperparameters and training procedures of the base methods. Code is publicly available at https://github.com/szq0214/Un-Mix.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

We revisit Semantic Scene Completion (SSC), a useful task to predict the semantic and occupancy representation of 3D scenes, in this paper. A number of methods for this task are always based on voxelized scene representations. Although voxel representations keep local structures of the scene, these methods suffer from heavy computation redundancy due to the existence of visible empty voxels when the network goes deeper. To address this dilemma, we propose our novel point-voxel aggregation network for this task. We first transfer the voxelized scenes to point clouds by removing these visible empty voxels and adopt a deep point stream to capture semantic information from the scene efficiently. Meanwhile, a light-weight voxel stream containing only two 3D convolution layers preserves local structures of the voxelized scenes. Furthermore, we design an anisotropic voxel aggregation operator to fuse the structure details from the voxel stream into the point stream, and a semantic-aware propagation module to enhance the up-sampling process in the point stream by semantic labels. We demonstrate that our model surpasses state-of-the-arts on two benchmarks by a large margin, with only the depth images as input.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

To enhance low-light images to normally-exposed ones is highly ill-posed, namely that the mapping relationship between them is one-to-many. Previous works based on the pixel-wise reconstruction losses and deterministic processes fail to capture the complex conditional distribution of normally exposed images, which results in improper brightness, residual noise, and artifacts. In this paper, we investigate to model this one-to-many relationship via a proposed normalizing flow model. An invertible network that takes the low-light images/features as the condition and learns to map the distribution of normally exposed images into a Gaussian distribution. In this way, the conditional distribution of the normally exposed images can be well modeled, and the enhancement process, i.e., the other inference direction of the invertible network, is equivalent to being constrained by a loss function that better describes the manifold structure of natural images during the training. The experimental results on the existing benchmark datasets show our method achieves better quantitative and qualitative results, obtaining better-exposed illumination, less noise and artifact, and richer colors.

poster session 5 @ red 2, poster session 12 @ red 2, oral session 12 @ red 2, poster session 5, poster session 12, oral session 12

Masked Language Models (MLMs) have shown superior performances in numerous downstream Natural Language Processing (NLP) tasks. Unfortunately, MLMs also demonstrate significantly worrying levels of social biases. We show that the previously proposed evaluation metrics for quantifying the social biases in MLMs are problematic due to the following reasons: (1) prediction accuracy of the masked tokens itself tend to be low in some MLMs, which leads to unreliable evaluation metrics, and (2) in most downstream NLP tasks, masks are not used; therefore prediction of the mask is not directly related to them, and (3) high-frequency words in the training data are masked more often, introducing noise due to this selection bias in the test cases. Therefore, we propose All Unmasked Likelihood (AUL), a bias evaluation measure that predicts \emph{all} tokens in a test case given the MLM embedding of the \emph{unmasked} input and AUL with Attention weights (AULA) to evaluate tokens based on their importance in a sentence. Our experimental results show that the proposed bias evaluation measures accurately detect different types of biases in MLMs, and unlike AUL and AULA, previously proposed measures for MLMs systematically overestimate the measured biases and are heavily influenced by the unmasked tokens in the context.

poster session 5 @ red 6, poster session 12 @ red 6, poster session 5, poster session 12

Advantage learning (AL) aims to improve the robustness of value-based reinforcement learning against estimation errors with action-gap-based regularization. Unfortunately, the method tends to be unstable in the case of function approximation. In this paper, we propose a simple variant of AL, named smoothing advantage learning (SAL), to alleviate this problem. The key to our method is to replace the original Bellman Optimal operator in AL with a smooth one so as to obtain more reliable estimation of the temporal difference target. We give a detailed account of the resulting action gap and the performance bound for approximate SAL. Further theoretical analysis reveals that the proposed value smoothing technique not only helps to stabilize the training procedure of AL by controlling the trade-off between convergence rate and the upper bound of the approximation errors, but is beneficial to increase the action gap between the optimal and sub-optimal action value as well.

poster session 6 @ blue 1, poster session 12 @ blue 1, poster session 6, poster session 12

This paper starts by revealing a surprising finding: without any learning, a randomly initialized CNN can localize objects surprisingly well. That is, a CNN has an inductive bias to naturally focus on objects, named as Tobias ("The object is at sight") in this paper. This empirical inductive bias is further analyzed and successfully applied to self-supervised learning (SSL). A CNN is encouraged to learn representations that focus on the foreground object, by transforming every image into various versions with different backgrounds, where the foreground and background separation is guided by Tobias. Experimental results show that the proposed Tobias significantly improves downstream tasks, especially for object detection. This paper also shows that Tobias has consistent improvements on training sets of different sizes, and is more resilient to changes in image augmentations.

poster session 5 @ red 3, poster session 12 @ red 3, oral session 12 @ red 3, poster session 5, poster session 12, oral session 12

In this paper, we propose a novel method to extend sequence-to-sequence models to accurately process sequences much longer than the ones used during training while being sample- and resource-efficient, supported by thorough experimentation. To investigate the effectiveness of our method, we apply it to the task of correcting documents already processed with Optical Character Recognition (OCR) systems using sequence-to-sequence models based on characters. We test our method on nine languages of the ICDAR 2019 competition on post-OCR text correction and achieve a new state-of-the-art performance in five of them. The strategy with the best performance involves splitting the input document in character n-grams and combining their individual corrections into the final output using a voting scheme that is equivalent to an ensemble of a large number of sequence models. We further investigate how to weigh the contributions from each one of the members of this ensemble. Our code for post-OCR correction is shared at https://github.com/jarobyte91/post_ocr_correction.

poster session 5 @ red 5, poster session 12 @ red 5, poster session 5, poster session 12

Pursuing accurate and robust recognizers has been a long-lasting goal for scene text recognition (STR) researchers. Recently, attention-based methods have demonstrated their effectiveness and achieved impressive results on public benchmarks. The attention mechanism enables models to recognize scene text with severe visual distortions by leveraging contextual information. However, recent studies revealed that the implicit over-reliance of context leads to catastrophic out-of-vocabulary performance. On the contrary to the superior accuracy of the seen text, models are prone to misrecognize unseen text even with good image quality. We propose a novel framework, Context-based contrastive learning (ConCLR), to alleviate this issue. Our proposed method first generates characters with different contexts via simple image concatenation operations and then optimizes contrastive loss on their embeddings. By pulling together clusters of identical characters within various contexts and pushing apart clusters of different characters in embedding space, ConCLR suppresses the side-effect of overfitting to specific contexts and learns a more robust representation. Experiments show that ConCLR significantly improves out-of-vocabulary generalization and achieves state-of-the-art performance on public benchmarks together with attention-based recognizers.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

Single image deraining is an important and challenging task for some downstream artificial intelligence applications such as video surveillance and self-driving systems. Most of the existing deep-learning-based methods constrain the network to generate derained images but few of them explore features from intermediate layers, different levels, and different modules which are beneficial for rain streaks removal. In this paper, we propose a high-order collaborative network with multi-scale compact constraints and a bidirectional scale-content similarity mining module to exploit features from deep networks externally and internally for rain streaks removal. Externally, we design a deraining framework with three sub-networks trained in a collaborative manner, where the bottom network transmits intermediate features to the middle network which also receives shallower rainy features from the top network and sends features back to the bottom network. Internally, we enforce multi-scale compact constraints on the intermediate layers of deep networks to learn useful features via a Laplacian pyramid. Further, we develop a bidirectional scale-content similarity mining module to explore features at different scales in a down-to-up and up-to-down manner. To improve the model performance on real-world images, we propose an online-update learning approach, which uses real-world rainy images to fine-tune the network and update the deraining results in a self-supervised manner. Extensive experiments demonstrate that our proposed method performs favorably against eleven state-of-the-art methods on five public synthetic datasets and one real-world dataset. The source code can be found at: \url{https://supercong94.wixsite.com/supercong94}.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

Dynamic convolution has achieved significant gain in performance and computational complexity, thanks to its powerful representation capability given limited filter number/layers. However, SOTA dynamic convolution operators are sensitive to input noises (\eg, Gaussian noise, shot noise, \etc) and lack sufficient spatial contextual information in filter generation. To alleviate this inherent weakness, we propose a lightweight and heterogeneous-structure (\ie, static and dynamic) operator, named \emph{Bi-volution}. On the one hand, Bi-volution is designed as a dual-branch structure to fully leverage complementary properties of static/dynamic convolution, which endows Bi-volution more robust properties and higher performance. On the other hand, the Spatial Augmented Kernel Generation module is proposed to improve the dynamic convolution, realizing the learning of spatial context information with negligible additional computational complexity. Extensive experiments illustrate that the ResNet-50 equipped with Bi-volution achieves a highly competitive boost in performance ($+2.8\%$ top-1 accuracy on ImageNet classification, $+2.4\%$ box AP and $+2.2\%$ mask AP on COCO detection and instance segmentation) while maintaining extremely low FLOPs (\ie, ResNet50$@2.7$ GFLOPs). Furthermore, our Bi-volution shows better robustness than dynamic convolution against various noise and input corruptions.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

Graph convolutional networks have been successfully applied to the task of event detection. However, existing works rely heavily on a fixed syntactic parse tree structure from an external parser. In addition, the information content extracted for aggregation is determined simply by the (syntactic) edge direction or type but irrespective of what semantics the vertices have, which is somewhat rigid. With this work, we propose a novel graph convolutional method that combines an adaptive graph generation technique and a multi-channel selection strategy. The adaptive graph generation technique enables the gradients to pass through the graph sampling layer by using the ST-Gumbel-Softmax trick. The multi-channel selection strategy allows two adjacent vertices to automatically determine which information channels to get through for information extraction and aggregation. The proposed method achieves the state-of-the-art performance on ACE2005 dataset.

poster session 5 @ red 5, poster session 12 @ red 5, poster session 5, poster session 12

Existing approaches for 2D pose estimation in videos often require a large number of dense annotations, which are costly and labor intensive to acquire. In this paper, we propose a semi-supervised REinforced MOtion Transformation nEtwork (REMOTE) to leverage a few labeled frames and temporal pose variations in videos, which enables effective learning of 2D pose estimation in sparsely annotated videos. Specifically, we introduce a Motion Transformer (MT) module to perform cross frame reconstruction, aiming to learn motion dynamic knowledge in videos. Besides, a novel reinforcement learning-based Frame Selection Agent (FSA) is designed within our framework, which is able to harness informative frame pairs on the fly to enhance the pose estimator under our cross reconstruction mechanism. We conduct extensive experiments that show the efficacy of our proposed REMOTE framework.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Slides are broadly used to present the research works and there are several studies on the problem of automatic slides generation. However, the lack of dataset hinders further research. In this paper, we construct a benchmark dataset for the problem of slides generation from scholarly papers. The dataset is fine-grained and consists of aligned pairs of single slide and specific region of a paper. Then we deploy several baseline models and conduct preliminary experiments. The results show that this task is challenging and awaits more exploration. The dataset and code will be released.

poster session 5 @ red 5, poster session 12 @ red 5, poster session 5, poster session 12

Geospatio-temporal data are pervasive across numerous application domains, including healthcare, environmental sciences, criminology, and urban computing. These rich datasets can be harnessed to predict extreme events such as disease outbreaks, flooding, crime spikes, traffic congestion, etc. However, since the extreme events are rare, predicting them is a hard problem. Statistical methods based on extreme value theory provide a systematic way for modeling the distribution of extreme values. In particular, the generalized Pareto distribution (GPD) is useful for modeling the distribution of excess values above a certain threshold. However, applying such methods to large-scale geospatio-temporal data is a challenge due to the difficulty in capturing the complex spatial relationships between extreme events at multiple locations. This paper presents a deep learning framework for long-term prediction of the distribution of extreme values at different locations. We highlight its computational challenges and present a novel framework that combines convolutional neural networks with deep set and GPD. We demonstrate the effectiveness of our approach on a real-world dataset for modeling extreme climate events.

poster session 6 @ blue 5, poster session 12 @ blue 5, oral session 6 @ blue 5, poster session 6, poster session 12, oral session 6

Symbolic search, using Binary Decision Diagrams (BDDs) to represent sets of states, is a competitive approach to optimal planning. Yet heuristic search in this context remains challenging. The many advances on admissible planning heuristics are not directly applicable, as they evaluate one state at a time. Indeed, progress using heuristic functions in symbolic search has been limited and even very informed heuristics have been shown to be detrimental. Here we show how this connection can be made stronger for LP-based potential heuristics. Our key observation is that, for this family of heuristic functions, the change of heuristic value induced by each operator can be precomputed. This facilitates their smooth integration into symbolic search. Our experiments show that this can pay off significantly: we establish a new state of the art in optimal symbolic planning.

poster session 3 @ blue 4, poster session 12 @ blue 4, oral session 12 @ blue 4, poster session 3, poster session 12, oral session 12

In a wide variety of applications including online advertising, contractual hiring, and wireless scheduling, the controller is constrained by a stringent budget constraint on the available resources, which are consumed in a random amount by each action, and a stochastic feasibility constraint that may impose important operational limitations on decision-making. In this work, we consider a general model to address such problems, where each action returns a random reward, cost, and penalty from an unknown joint distribution, and the decision-maker aims to maximize the total reward under a budget constraint $B$ on the total cost and a stochastic constraint on the time-average penalty. We propose a novel low-complexity algorithm based on Lyapunov optimization methodology, named ${\tt LyOn}$, and prove that for $K$ arms it achieves $O(\sqrt{KB\log B})$ regret and zero constraint-violation when $B$ is sufficiently large. The low computational cost and sharp performance bounds of ${\tt LyOn}$ suggest that Lyapunov-based algorithm design methodology can be effective in solving constrained bandit optimization problems.

poster session 1 @ blue 3, poster session 12 @ blue 3, oral session 1 @ blue 3, poster session 1, poster session 12, oral session 1

Social media, blogs, and online articles are instant sources of news for internet users globally. But due to their unmoderated nature, a significant percentage of these texts are fake news or rumors. Their deceptive nature and ability to propagate instantly can have an adverse effect on society. In this work, we hypothesize that legitimacy of news has a correlation with its emotion, and propose a multi-task framework predicting both the emotion and legitimacy of news. Experimental results verify that our multi-task models outperform their single-task counterparts in terms of accuracy.

poster session 5 @ blue 4, poster session 12 @ blue 4, poster session 5, poster session 12

Within the field of face recognition (FR), it is widely accepted that the key objective is to optimize the entire feature space in the training process and acquire robust feature representations. However, most real-world FR systems tend to operate at a pre-defined False Accept Rate (FAR), and the corresponding True Accept Rate (TAR) represents the performance of the FR systems, which indicates that the optimization on the pre-defined FAR is more meaningful and important in the practical evaluation process. In this paper, we call the predefined FAR as Anchor FAR, and we argue that the existing FR loss functions cannot guarantee the optimal TAR under the Anchor FAR, which impedes further improvements of FR systems. To this end, we propose AnchorFace to bridge the aforementioned gap between the training and practical evaluation process for FR. Given the Anchor FAR, AnchorFace can boost the performance of FR systems by directly optimizing the non-differentiable FR evaluation metrics. Specifically, in AnchorFace, we first calculate the similarities of the positive and negative pairs based on both the features of the current batch and the stored features in the maintained online- updating set, and then generate the differentiable TAR loss and FAR loss using a soften strategy. Our AnchorFace can be readily integrated into most existing FR loss functions, and extensive experimental results on multiple benchmark datasets demonstrate the effectiveness of AnchorFace.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Pre-training language models (LMs) on large-scale unlabeled text data makes the model much easier to achieve exceptional downstream performance than their counterparts directly trained on the downstream tasks. In this work, we study what specific traits in the pre-training data, other than the semantics, make a pre-trained LM superior to their counterparts trained from scratch on downstream tasks. We propose to use artificially constructed datasets as the pre-training data to exclude the effect of semantics, and further control what characteristics the pre-training corpora have. By fine-tuning the pre-trained models on GLUE benchmark, we can learn how beneficial it is to transfer the knowledge from the model trained on the dataset possessing that specific trait. We define and discuss three different characteristics in the artificial dataset: 1) matching the token's uni-gram or bi-gram distribution between pre-training and downstream fine-tuning, 2) the presence of the explicit dependencies among the tokens in a sequence, 3) the length of the implicit dependencies among the tokens in a sequence. Our experiments show that the explicit dependencies in the sequences of the pre-training data are critical to the downstream performance. Our results also reveal that models achieve better downstream performance when pre-trained on a dataset with a longer range of implicit dependencies. Based on our analysis, we find that models pre-trained with artificial datasets are prone to learn spurious correlation in downstream tasks. Our work reveals that even if the LMs are not pre-trained on natural language, they still gain transferability on certain human language downstream tasks once the LMs learn to model the token dependencies in the sequences. This result helps us understand the exceptional transferability of pre-trained LMs.

poster session 3 @ red 4, poster session 12 @ red 4, oral session 12 @ red 4, poster session 3, poster session 12, oral session 12

Human-Object Interaction (HOI) detection is an essential task to understand human-centric images from a fine-grained perspective. Although end-to-end HOI detection models thrive, their paradigm of parallel human/object detection and verb class prediction loses two-stage methods' merit: object-guided hierarchy. The object in one HOI triplet gives direct clues to the verb to be predicted. In this paper, we aim to boost end-to-end models with object-guided statistical priors. Specifically, We propose to utilize a Verb Semantic Model (VSM) and use semantic aggregation to profit from this object-guided hierarchy. Similarity KL (SKL) loss is proposed to optimize VSM to align with the HOI dataset's priors. To overcome the static semantic embedding problem, we propose to generate cross-modality-aware visual and semantic features by Cross-Modal Calibration (CMC). The above modules combined composes Object-guided Cross-modal Calibration Network (OCN). Experiments conducted on two popular HOI detection benchmarks demonstrate the significance of incorporating the statistical prior knowledge and produce state-of-the-art performances. More detailed analysis indicates proposed modules serve as a stronger verb predictor and a more superior method of utilizing prior knowledge. The codes are available at https://github.com/JacobYuan7/OCN-HOI-Benchmark.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

Object detection in high-resolution satellite imagery is emerging as a scalable alternative to on-the-ground survey data collection in many environmental and socioeconomic monitoring applications. However, performing object detection over large geographies can still be prohibitively expensive due to the high cost of purchasing imagery and compute. Inspired by traditional survey data collection strategies, we propose an approach to estimate object count statistics over large geographies through sampling. Given a cost budget, our method selects a small number of representative areas by sampling from a learnable proposal distribution. Using importance sampling, we are able to accurately estimate object counts after processing only a small fraction of the images compared to an exhaustive approach. We show empirically that the proposed framework achieves strong performance on estimating the number of buildings in the United States and Africa, cars in Kenya, brick kilns in Bangladesh, and swimming pools in the U.S., while requiring as few as 0.01% of satellite images compared to an exhaustive approach.

poster session 3 @ red 6, poster session 12 @ red 6, oral session 12 @ red 6, poster session 3, poster session 12, oral session 12

High-performance traffic flow prediction model designing, a core technology of Intelligent Transportation System, is a long-standing but still challenging task for industrial and academic communities. The lack of integration between physical principle and data-driven model is an important reason for limiting the development of this field. In the literature, the physics-based methods can usually provide a clear interpretation of the dynamic process of traffic flow systems but are with limited accuracy, while data-driven methods, especially deep learning with black-box structures, can achieve improved performance but can not be fully trusted due to lack of a reasonable physical basis. To bridge the gap between purely data-driven and physics-driven approaches, we propose a physics-guided deep learning model named Spatio-Temporal Differential Equation Network (STDEN), which casts the physical mechanism of traffic flow dynamics into a deep neural network framework. Specifically, we assume the traffic flow on road networks is driven by a latent potential energy field (like water flows are driven by the gravity field), and model the spatio-temporal dynamic process of the potential energy field as a differential equation network. STDEN absorbs both the performance advantage of data-driven models and the interpretability of physics-based models, so is named a physics-guided prediction model. Experiments on three real-world traffic datasets in Beijing show that our model outperforms state-of-the-art baselines by a significant margin. A case study further verifies that STDEN can capture the mechanism of urban traffic and generate accurate predictions with physical meaning. The proposed framework of differential equation network modeling may also cast light on other similar applications.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

Running Convolutional Neural Networks (CNNs) on resource strictly restricted IoT devices, e.g. microcontroller units (MCU) and field programmable gate array (FPGA), is an important and challenging topic: Tiny machine learning (TinyML). To meet very limited storage and power requirements of IoT devices, it is essential to reduce the parameters of CNN models as much as possible. In this paper, we aim at developing an efficient tiny model with only hundreds of KB parameters. Toward this end, we firstly design a parameter-efficient tiny architecture by introducing dense linear depthwise block. Then, a novel adaptive scale quantization (ASQ) method is proposed for further quantizing tiny models in aggressive low-bit while retaining the accuracy. With the optimized architecture and 4-bit ASQ, we present a family of ultra-lightweight networks, named EtinyNet, that achieves 57.0% ImageNet top-1 accuracy with an extremely tiny model size of 340KB. When deployed on an off-the-shelf commercial microcontroller for object detection tasks, EtinyNet achieves state-of-the-art 56.4% mAP on Pascal VOC. Furthermore, the experimental results on Xilinx compact FPGA indicate that EtinyNet achieves prominent low power of 620mW, about 5.6x lower than existing FPGA designs.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

Universal domain adaptation (UniDA) aims to transfer the knowledge learned from a labeled source domain to an unlabeled target domain without any constraints on the label sets. However, domain shift and category shift make UniDA extremely challenging, mainly attributed to the requirement of identifying both shared “known” samples and private “unknown” samples. Previous methods barely exploit the intrinsic manifold structure relationship between two domains for feature alignment, and they rely on the softmax-based scores with class competition nature to detect underlying “unknown” samples. Therefore, in this paper, we propose a novel eviden\textbf{T}ial \textbf{N}eighborhood con\textbf{T}rastive learning framework called TNT to address these issues. Specifically, TNT first proposes a new domain alignment principle: semantically consistent samples should be geometrically adjacent to each other, whether within or across domains. From this criterion, a cross-domain multi-sample contrastive loss based on mutual nearest neighbors is designed to achieve common category matching and private category separation. Second, toward accurate “unknown” sample detection, TNT introduces a class competition-free uncertainty score from the perspective of evidential deep learning. Instead of setting a single threshold, TNT learns a category-aware heterogeneous threshold vector to reject diverse “unknown” samples. Extensive experiments on three benchmarks demonstrate that TNT significantly outperforms previous state-of-the-art UniDA methods.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

3D hand pose estimation from single depth is a fundamental problem in computer vision, and has wide applications. However, the existing methods still can not achieve satisfactory hand pose estimation results due to view variation and occlusion of human hand. In this paper, we propose a new virtual view selection and fusion module for 3D hand pose estimation from single depth. We propose to automatically select multiple virtual viewpoints for pose estimation and fuse the results of all and find this empirically delivers accurate and robust pose estimation. In order to select most effective virtual views for pose fusion, we evaluate the virtual views based on the confidence of virtual views using a light-weight network via network distillation. Experiments on three main benchmark datasets including NYU, ICVL and Hands2019 demonstrate that our method outperforms the state-of-the-arts on NYU and ICVL, and achieves very competitive performance on Hands2019-Task1, and our proposed virtual view selection and fusion module is both effective for 3D hand pose estimation. The code is available in project webpage https://github.com/iscas3dv/handpose-virtualview.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Structural Equation/Causal Models (SEMs/SCMs) are widely used in epidemiology and social sciences to identify and analyze the average treatment effect (ATE) and conditional ATE (CATE). Traditional causal effect estimation methods such as Inverse Probability Weighting (IPW) and more recently Regression-With-Residuals (RWR) are widely used - as they avoid the challenging task of identifying the SCM parameters - to estimate ATE and CATE. However, much work remains before traditional estimation methods can be used for counterfactual inference, and for the benefit of Personalized Public Policy Analysis (P$^3$A) in the social sciences. While doctors rely on personalized medicine to tailor treatments to patients in laboratory settings (relatively closed systems), P$^3$A draws inspiration from such tailoring but adapts it for open social systems. In this article, we develop a method for counterfactual inference that we name causal-Graphical Normalizing Flow (c-GNF), facilitating P$^3$A. A major advantage of c-GNF is that it suits the open system in which P$^3$A is conducted. First, we show how c-GNF captures the underlying SCM without making any assumption about functional forms. This capturing capability is enabled by the deep neural networks that model the underlying SCM via observational data likelihood maximization using gradient descent. Second, we propose a novel dequantization trick to deal with discrete variables, which is a limitation of normalizing flows in general. Third, we demonstrate in experiments that c-GNF performs on-par with IPW and RWR in terms of bias and variance for estimating the ATE, when the true functional forms are known, and better when they are unknown. Fourth and most importantly, we conduct counterfactual inference with c-GNFs, demonstrating promising empirical performance. Because IPW and RWR, like other traditional methods, lack the capability of counterfactual inference, c-GNFs will likely play a major role in tailoring personalized treatment, facilitating P$^3$A, optimizing social interventions - in contrast to the current `one-size-fits-all' approach of existing methods.

poster session 3 @ red 6, poster session 12 @ red 6, oral session 12 @ red 6, poster session 3, poster session 12, oral session 12

The hierarchical interaction between the actor and critic in actor-critic based reinforcement learning algorithms naturally lends itself to a game-theoretic interpretation. We adopt this viewpoint and model the actor and critic interaction as a two-player general-sum game with a leader-follower structure known as a Stackelberg game. Given this abstraction, we propose a meta-framework for Stackelberg actor-critic algorithms where the leader player follows the total derivative of its objective instead of the usual individual gradient. From a theoretical standpoint, we develop a policy gradient theorem for the refined update and provide a local convergence guarantee for the Stackelberg actor-critic algorithms to a local Stackelberg equilibrium. From an empirical standpoint, we demonstrate via simple examples that the learning dynamics we study mitigate cycling and accelerate convergence compared to the usual gradient dynamics given cost structures induced by actor-critic formulations. Finally, extensive experiments on OpenAI gym environments show that Stackelberg actor-critic algorithms always perform at least as well and often significantly outperform the standard actor-critic algorithm counterparts.

poster session 6 @ blue 1, poster session 12 @ blue 1, poster session 6, poster session 12

Conditional density estimation (CDE) is the task of estimating the probability of an event conditioned on some inputs. A neural network (NN) can also be used to compute the output distribution for continuous-domain, which can be viewed as an extension of regression task. Nevertheless, it is difficult to explicitly approximate a distribution without knowing the information of its general form a priori. In order to fit an arbitrary conditional distribution, discretizing the continuous domain into bins is an effective strategy, as long as we have sufficiently narrow bins and very large data. However, collecting enough data is often hard to reach and falls far short of that ideal in many circumstances, especially in multivariate CDE for the curse of dimensionality. In this paper, we demonstrate the benefits of modeling free-form conditional distributions using a deconvolution-based neural net framework, coping with data deficiency problems in discretization. It has the advantage of being flexible but also takes advantage of the hierarchical smoothness offered by the deconvolution layers. We compare our method to a number of other density-estimation approaches and show that our Deconvolutional Density Network (DDN) outperforms the competing methods on many univariate and multivariate tasks. The code of DDN is available at https://github.com/NBICLAB/DDN

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

Electricity forecasting has important implications for the key decisions in modern electricity systems, ranging from power generation, transmission, distribution and so on. In the literature, traditional statistic approaches, machine-learning methods and deep learning (e.g., recurrent neural network) based models are utilized to model the trends and patterns in electricity time-series data. However, they are restricted either by their deterministic forms or by independence in probabilistic assumptions -- thereby neglecting the uncertainty or significant correlations between distributions of electricity data. Ignoring these, in turn, may yield error accumulation, especially when relying on historical data and aiming at multi-step prediction. To overcome these, we propose a novel method named Probabilistic Electricity Forecasting (PrEF) by proposing a non-linear neural state space model (SSM) and incorporating copula-augmented mechanism into that, which can learn uncertainty-dependencies knowledge and understand interactive relationships between various factors from large-scale electricity time-series data. Our method distinguishes itself from existing models by its traceable inference procedure and its capability of providing high-quality probabilistic distribution predictions. Extensive experiments on two real-world electricity datasets demonstrate that our method consistently outperforms the alternatives.

poster session 5 @ red 6, poster session 12 @ red 6, poster session 5, poster session 12

The vast amount of data captured by robots in large-scale environments brings the computing and storage bottlenecks to the typical methods of modeling the spaces the robots travel in. In order to efficiently construct a compact terrain model from uncertain, incomplete point cloud data of large-scale environments, in this paper, we first propose a novel feature mapping method, named random mapping, based on the fast random construction of base functions, which can efficiently project the messy points in the low-dimensional space into the high-dimensional space where the points are approximately linearly distributed. Then, in this mapped space, we propose to learn a continuous linear regression model to represent the terrain. We show that this method can model the environments in much less computation time, memory consumption and access time with high accuracy. Furthermore, the models possess the generalization capabilities comparable to the performances on the training set, and its inference accuracy gradually increases as the random mapping dimension increases. To better solve the large-scale environmental modeling problem, we adopt the idea of parallel computing to train the models. This strategy greatly reduce the wall-clock time of calculation without losing much accuracy. Experiments show the effectiveness of random mapping method and effects of some important parameters on its performances. Moreover, we evaluate the proposed terrain modeling method based on the random mapping method and compare its performances with popular typical methods and state-of-art methods.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

We present Laneformer, a conceptually simple yet powerful transformer-based architecture tailored for lane detection that is a long-standing research topic for visual perception in autonomous driving. The dominant paradigms rely on purely CNN-based architectures which often fail in incorporating relations of long-range lane points and global contexts induced by surrounding objects (e.g., pedestrians, vehicles). Inspired by recent advances of the transformer encoder-decoder architecture in various vision tasks, we move forwards to design a new end-to-end Laneformer architecture that revolutionizes the conventional transformers into better capturing the shape and semantic characteristics of lanes, with minimal overhead in latency. First, coupling with deformable pixel-wise selfattention in the encoder, Laneformer presents two new row and column self-attention operations to efficiently mine point context along with the lane shapes. Second, motivated by the appearing objects would affect the decision of predicting lane segments, Laneformer further includes the detected object instances as extra inputs of multi-head attention blocks in the encoder and decoder to facilitate the lane point detection by sensing semantic contexts. Specifically, the bounding box locations of objects are added into Key module to provide interaction with each pixel and query while the ROI-aligned features are inserted into Value module. Extensive experiments demonstrate our Laneformer achieves state-of-the-art performances on CULane benchmark, in terms of 77.1% F1 score. We hope our simple and effective Laneformer will serve as a strong baseline for future research in self-attention models for lane detection.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

Data-driven methods have achieved notable performance on intent detection, which is a task to comprehend user queries. Nonetheless, they are controversial for over-confident predictions. In some scenarios, users do not only care about the accuracy but also the confidence of model. Unfortunately, mainstream neural networks are poorly calibrated, with a large gap between accuracy and confidence. To handle this problem defined as confidence calibration, we propose a model using the hyperspherical space and rebalanced accuracy-uncertainty loss. Specifically, we project the label vector onto hyperspherical space uniformly to generate a dense label representation matrix, which mitigates over-confident predictions due to overfitting sparse one-hot label matrix. Besides, we rebalance samples of different accuracy and uncertainty to better guide model training. Experiments on the open datasets verify that our model outperforms the existing calibration methods and achieves a significant improvement on the calibration metric.

poster session 3 @ red 4, poster session 12 @ red 4, poster session 3, poster session 12

Recent works have shown explainability and robustness are two crucial ingredients of trustworthy and reliable text classification. However, previous works usually address one of two aspects: i) how to extract accurate rationales for explainability while being beneficial to prediction; ii) how to make the predictive model robust to different types of adversarial attacks. Intuitively, a model that produces helpful explanations should be more robust against adversarial attacks, because we cannot trust the model that outputs explanations but changes its prediction under small perturbations. To this end, we propose a joint classification and rationale extraction model named AT-BMC. It includes two key mechanisms: mixed Adversarial Training (AT) is designed to use various perturbations in discrete and embedding space to improve the model’s robustness, and Boundary Match Constraint (BMC) helps to locate rationales more precisely with the guidance of boundary information. Performances on benchmark datasets demonstrate that the proposed AT-BMC outperforms baselines on both classification and rationale extraction by a large margin. Robustness analysis shows that the proposed AT-BMC decreases the attack success rate effectively by up to 69%. The results indicate that there are connections between robust models and better explanations.

poster session 3 @ red 4, poster session 12 @ red 4, oral session 3 @ red 4, poster session 3, poster session 12, oral session 3

$k$-means and $k$-median clustering are powerful unsupervised machine learning techniques. However, due to complicated dependences on all the features, it is challenging to interpret the resulting cluster assignments. Moshkovitz, Dasgupta, Rashtchian, and Frost [ICML 2020] proposed an elegant model of explainable $k$-means and $k$-median clustering. In this model, a decision tree with $k$ leaves provides a straightforward characterization of the data set into clusters. We study two natural algorithmic questions about explainable clustering. (1) For a given clustering, how to find the ``best explanation'' by using a decision tree with $k$ leaves? (2) For a given set of points, how to find a decision tree with $k$ leaves minimizing the $k$-means/median objective of the resulting explainable clustering? To address the first question, we introduce a new model of explainable clustering. Our model, inspired by the notion of outliers in robust statistics, is the following. We are seeking a small number of points (outliers) whose removal makes the existing clustering well-explainable. For addressing the second question, we initiate the study of the model of Moshkovitz et al. from the perspective of multivariate complexity. Our rigorous algorithmic analysis sheds some light on the influence of parameters like the input size, dimension of the data, the number of outliers, the number of clusters, and the approximation ratio, on the computational complexity of explainable clustering.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

Group Lasso is an important sparse regression method in machine learning which encourages selecting key explanatory factors in a grouped manner because of the use of $\ell_{2,1}$-norm. In real-world learning tasks, some chunks of data would be added into or removed from the training set in sequence due to the existence of new or obsolete historical data, which is normally called dynamic or lifelong learning scenario. However, most of existing algorithms of group Lasso are limited to offline updating, and only one is online algorithm which can only handle newly added samples inexactly. Due to the complexity of $\ell_{2,1}$-norm, how to achieve accurate chunk incremental and decremental learning efficiently for group Lasso is still an open question. To address this challenging problem, in this paper, we propose a novel accurate dynamic updating algorithm for group Lasso by utilizing the technique of Ordinary Differential Equations (ODEs), which can incorporate or eliminate a chunk of samples from original training set without retraining the model from scratch. Specifically, we introduce a new formulation to reparameterize the adjustment procedures of chunk incremental and decremental learning simultaneously. Based on the new formulation, we propose a path following algorithm for group Lasso regarding to the adjustment parameter. Importantly, we prove that our path following algorithm can exactly track the piecewise smooth solutions thanks to the technique of ODEs, so that the accurate chunk incremental and decremental learning can be achieved. Extensive experimental results not only confirm the effectiveness of proposed algorithm for the chunk incremental and decremental learning, but also validate its efficiency compared to the existing offline and online algorithms.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

Recommender systems are facing scrutiny because of their growing impact on the opportunities we have access to. Current audits for fairness are limited to coarse-grained parity assessments at the level of sensitive groups. We propose to audit for \emph{envy-freeness}, a more granular criterion aligned with individual preferences: every user should prefer their recommendations to those of other users. Since auditing for envy requires to estimate the preferences of users beyond their existing recommendations, we cast the audit as a new pure exploration problem in multi-armed bandits. We propose a sample-efficient algorithm with theoretical guarantees that it does not deteriorate user experience. We also study the trade-offs achieved on real-world recommendation datasets.

poster session 1 @ blue 3, poster session 12 @ blue 3, oral session 12 @ blue 3, poster session 1, poster session 12, oral session 12

The rapid development of facial manipulation techniques has aroused public concerns in recent years. Existing deepfake video detection approaches attempt to capture the discriminative features between real and fake faces based on temporal modelling. However, these works impose supervisions on sparsely sampled video frames but overlook the local motions among adjacent frames, which instead encode rich inconsistency information that can serve as an effcient indicator for DeepFake video detection. To mitigate this issue, we delves into the local motion and propose a novel sampling unit named $snippet$ which contains a few successive videos frames for local temporal inconsistency learning. Moreover, we elaborately design an Intra-Snippet Inconsistency Module (Intra-SIM) and an Inter-Snippet Interaction Module (Inter-SIM) to establish a dynamic inconsistency modelling framework. Specifically, the Intra-SIM applies bi-directional temporal difference operations and a learnable convolution kernel to mine the short-term motions within each snippet. The Inter-SIM is then devised to promote the cross-snippet information interaction to form global representations. The Intra-SIM and Inter-SIM work in an alternate manner and can be plugged into existing 2D CNNs. Our method outperforms the state-of-the-art competitors on four popular benchmark dataset, \emph{i.e.,} FaceForensics++, Celeb-DF, DFDC and WildDeepfake. Besides, extensive experiments and visualizations are also presented to further illustrate its effectiveness.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

How to obtain good value estimation is a critical problem in Reinforcement Learning (RL). Current value estimation methods in continuous control, such as DDPG and TD3, suffer from unnecessary over- or under- estimation. In this paper, we explore the potential of double actors, which has been neglected for a long time, for better value estimation in the continuous setting. First, we interestingly find that double actors improve the exploration ability of the agent. Next, we uncover the bias alleviation property of double actors in handling overestimation with single critic, and underestimation with double critics respectively. Finally, to mitigate the potentially pessimistic value estimate in double critics, we propose to regularize the critics under double actors architecture. Together, we present Double Actors Regularized Critics (DARC) algorithm. Extensive experiments on challenging continuous control benchmarks, MuJoCo and PyBullet, show that DARC significantly outperforms current baselines with higher average return and better sample efficiency.

poster session 6 @ blue 1, poster session 12 @ blue 1, oral session 12 @ blue 1, poster session 6, poster session 12, oral session 12

In real world, training data for person re-identification (Re-ID) is collected discretely with spatial and temporal variations, which requires a model to incrementally learn new knowledge without forgetting old knowledge. This problem is called lifelong person re-identification (LReID). Variations of illumination and background for images of each task exhibit task-specific image style and lead to task-wise domain gap. In addition to missing data from the old tasks, task-wise domain gap is a key factor for catastrophic forgetting in LReID, which is ignored in existing approaches for LReID. The model tends to learn task-specific knowledge with task-wise domain gap, which results in stability and plasticity dilemma. To overcome this problem, we cast LReID as a domain adaptation problem and propose a pseudo task knowledge preservation framework to alleviate the domain gap. Our framework is based on a pseudo task transformation module which maps the features of the new task into the feature space of the old tasks to complement the limited saved exemplars of the old tasks. With extra transformed features in the task-specific feature space, we propose a task-specific domain consistency loss to implicitly alleviate the task-wise domain gap for learning task-shared knowledge instead of task-specific one. Furthermore, to guide knowledge preservation with the feature distributions of the old tasks, we propose to preserve knowledge on extra pseudo tasks which jointly distills knowledge and discriminates identity, in order to achieve a better trade-off between stability and plasticity for lifelong learning with task-wise domain gap. Extensive experiments demonstrate the superiority of our method as compared with the state-of-the-art lifelong learning and LReID methods.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Offline reinforcement learning aims to maximize the expected cumulative rewards with a fixed collection of data. The basic principle of current offline reinforcement learning methods is to restrict the policy to the offline dataset action space. However, they ignore the case where the dataset's trajectories fail to cover the state space completely. Especially, when the dataset's size is limited, it is likely that the agent would encounter unseen states during test time. Prior policy-constrained methods are incapable of correcting the state deviation, and may lead the agent to its unexpected regions further. In this paper, we propose the state deviation correction (SDC) method to constrain the policy's induced state distribution by penalizing the out-of-distribution states which might appear during the test period. We first perturb the states sampled from the logged dataset, then simulate noisy next states on the basis of a dynamics model and the policy. We then train the policy to minimize the distances between the noisy next states and the offline dataset. In this manner, we allow the trained policy to guide the agent to its familiar regions. Experimental results demonstrate that our proposed method is competitive with the state-of-the-art methods in a GridWorld setup, offline Mujoco control suite, and a modified offline Mujoco dataset with a finite number of valuable samples.

poster session 6 @ blue 1, poster session 12 @ blue 1, poster session 6, poster session 12

Image captioning has emerged as an interesting research field in recent years due to its broad application scenarios. The traditional paradigm of image captioning relies on paired image-caption datasets to train the model in a supervised manner. However, creating such paired datasets for every target language is prohibitively expensive, which hinders the extensibility of captioning technology and deprives a large part of the world population of its benefit. In this work, we present a novel unpaired cross-lingual method to generate image captions without relying on any caption corpus in the source or the target language. Specifically, our method consists of two phases: (1) a cross-lingual auto-encoding process, which utilizing a sentence parallel (bitext) corpus to learn the mapping from the source to the target language in the scene graph encoding space and decode sentences in the target language, and (2) a cross-modal unsupervised feature mapping, which seeks to map the encoded scene graph features from image modality to language modality. We verify the effectiveness of our proposed method on the Chinese image caption generation task. The comparisons against several existing methods demonstrate the effectiveness of our approach.

poster session 6 @ blue 2, poster session 12 @ blue 2, oral session 6 @ blue 2, poster session 6, poster session 12, oral session 6

Improving sample efficiency has been a longstanding goal in reinforcement learning. This paper proposes the $\mathtt{VRMPO}$ algorithm: a sample efficient policy gradient method with stochastic mirror descent. In $\mathtt{VRMPO}$, a novel variance-reduced policy gradient estimator is presented to improve sample efficiency. Furthermore, we prove that our $\mathtt{VRMPO}$ needs only $\mathcal{O}(\epsilon^{-3})$ sample trajectories to achieve an $\epsilon$-approximate first-order stationary point, which matches the best sample complexity. The extensive experimental results demonstrate that our algorithm outperforms the state-of-the-art policy gradient methods in various settings.

poster session 6 @ blue 1, poster session 12 @ blue 1, poster session 6, poster session 12

The task of Grammatical Error Correction (GEC) has received remarkable attention with wide applications in Natural Language Processing (NLP) in recent years. While one of the key principles of GEC is to keep the correct parts unchanged and avoid over-correction, previous sequence-to-sequence (seq2seq) models generate results from scratch, which are not guaranteed to follow the original sentence structure and may suffer from the over-correction problem. In the meantime, the recently proposed sequence tagging models can overcome the over-correction problem by only generating edit operations, but are conditioned on human designed language-specific tagging labels. In this paper, we combine the pros and alleviate the cons of both models by proposing a novel Sequence-to-Action (S2A) module. The S2A module jointly takes the source and target sentences as input, and is able to automatically generate a token-level action sequence before predicting each token, where each action is generated from three choices named \texttt{SKIP}, \texttt{COPY} and \texttt{GEN}erate. Then the actions are fused with the basic seq2seq framework to provide final predictions. We conduct experiments on the benchmark datasets of both English and Chinese GEC tasks. Our model consistently outperforms the seq2seq baselines, while being able to significantly alleviate the over-correction problem as well as holding better generality and diversity in the generation results compared to the sequence tagging models.

poster session 5 @ red 5, poster session 12 @ red 5, poster session 5, poster session 12

The prosperity of mobile and financial technologies has bred and expanded various kinds of financial products to a broader scope of people, which contributes to financial inclusion. It brings non-trivial social benefits of diminishing financial inequality. However, the technical challenges in individual financial risk evaluation exacerbated by the unforeseen user characteristic distribution and limited credit history of new users, as well as the inexperience of newly-entered companies in handling complex data and obtaining accurate labels, impede further promotion of financial inclusion. To tackle these challenges, this paper develops a novel transfer learning algorithm (i.e., TransBoost) that combines the merits of tree-based models and kernel methods. The TransBoost is designed with a parallel tree structure and efficient weights updating mechanism with theoretical guarantee, which enables it to excel in tackling real-world data with high dimensional features and sparsity in O(n) time complexity. We conduct extensive experiments on two public datasets and a unique largescale dataset from Tencent Mobile Payment. The results show that the TransBoost outperforms other state-of-the- art benchmark transfer learning algorithms in terms of prediction accuracy with superior efficiency, demonstrate stronger robustness to data sparsity, and provide meaningful model interpretation. Besides, given a financial risk level, the TransBoost enables financial service providers to serve the largest number of users including those who would otherwise be excluded by other algorithms. That is, the TransBoost improves financial inclusion.

poster session 3 @ red 6, poster session 12 @ red 6, poster session 3, poster session 12

State of the art neural methods for open information extraction (OpenIE) usually extract triplets (or tuples) iteratively in an autoregressive or predicate-based manner in order not to produce duplicates. In this work, we propose a different approach to the problem that can be equally or more successful. Namely, we present a novel single-pass method for OpenIE inspired by object detection algorithms from computer vision. We use an order-agnostic loss based on bipartite matching that forces unique predictions and a Transformer-based encoder-only architecture for sequence labeling. The proposed approach is faster and shows superior or similar performance in comparison with state of the art models on standard benchmarks in terms of both quality metrics and inference time. Our model sets the new state of the art performance of 67.7% F1 on CaRB evaluated as OIE2016 while being 3.35x faster at inference than previous state of the art. We also evaluate the multilingual version of~our model in~the zero-shot setting for two languages and introduce a~strategy for generating synthetic multilingual data to fine-tune the model for each specific language. In this setting, we show performance improvement of 15\% on multilingual Re-OIE2016, reaching 75% F1 for both Portuguese and Spanish languages. Code and models are available at https://github.com/sberbank-ai/DetIE.

poster session 3 @ red 4, poster session 12 @ red 4, poster session 3, poster session 12

A riddle is a question or statement with double or veiled meanings, followed by an unexpected answer. Solving riddle is a challenging task for both machine and human, testing the capability of understanding figurative, creative natural language and reasoning with commonsense knowledge. We introduce BiRdQA, a bilingual multiple-choice question answering dataset with 6614 English riddles and 8751 Chinese riddles. For each riddle-answer pair, we provide four distractors with additional information from Wikipedia. The distractors are automatically generated at scale with minimal bias. Existing monolingual and multilingual QA models fail to perform well on our dataset, indicating that there is a long way to go before machine can beat human on solving tricky riddles. The dataset has been released to the community.

poster session 3 @ red 4, poster session 12 @ red 4, poster session 3, poster session 12

Volitionality and subject animacy are fundamental and closely related properties of events. Their classification, however, is challenging because it requires contextual text understanding and a huge amount of labeled data. This paper proposes a novel method that jointly learns volitionality and subject animacy at a low cost, heuristically labeling events in a raw corpus. Volitionality labels are assigned using a small lexicon of volitional and non-volitional adverbs such as deliberately and accidentally; subject animacy labels are assigned using a list of animate and inanimate nouns obtained from ontological knowledge. Since our labeling method assigns labels only to a biased set of events, a classifier is trained with regularization to take into account the property. This paper explores the following two approaches: bias reduction and adversarial representation learning. In bias reduction, the words used for labeling are regarded as bias that should not be over-exploited to make predictions, and their estimated contribution towards predictions is penalized. In adversarial representation learning, the classifier is given unlabeled events as well and makes their latent representations closer to labeled events' ones in an adversarial manner while learning classification on labeled events. We conduct experiments with crowdsourced gold data in Japanese and English and show that our method effectively learns volitionality and subject animacy without manually labeled data.

poster session 3 @ red 4, poster session 12 @ red 4, poster session 3, poster session 12

The daily practice of sharing images on social media raises a severe issue about privacy leakage. To address the issue, privacy-leaking image detection is studied recently, with the goal to automatically identify images that may leak privacy. Recent advance on this task benefits from focusing on crucial objects via pretrained object detectors and modeling their correlation. However, these methods have two limitations: 1) they neglect other important elements like scenes, textures, and objects beyond the capacity of pretrained object detectors. 2) the correlation among objects is fixed, but a fixed correlation is not appropriate for all the images. To overcome the limitations, we propose the Dynamic Region-Aware Graph Convolutional Network (DRAG) that dynamically finds out crucial regions including objects and other important elements, and model their correlation adaptively for each input image. To find out crucial regions, we cluster spatially-correlated feature channels into several region-aware feature maps. Furthermore, we dynamically model the correlation with the self-attention mechanism and explore the interaction among the regions with a graph convolutional network. The DRAG achieved an accuracy of 87% on the largest dataset for privacy-leaking image detection, which is 10 percentage points higher than the state of the art. The further case study demonstrates that it found out crucial regions containing not only objects but other important elements like textures. The code and more details are in https://github.com/guang-yanng/DRAG.

poster session 5 @ red 6, poster session 12 @ red 6, oral session 5 @ red 6, poster session 5, poster session 12, oral session 5

Distributed Constraint Optimization Problems (DCOPs) are an important subclass of combinatorial optimization problems, where information and controls are distributed among multiple autonomous agents. Previously, Machine Learning (ML) has been largely applied to solve combinatorial optimization problems by learning effective heuristics. However, existing ML-based heuristic methods are often not generalizable to different search algorithms. Most importantly, these methods usually require full knowledge about the problems to be solved, which are not suitable for distributed settings where centralization is not realistic due to geographical limitations or privacy concerns. To address the generality issue, we propose a novel directed acyclic graph representation schema for DCOPs and leverage the Graph Attention Networks (GATs) to embed graph representations. Our model, GAT-PCM, is then pretrained with optimally labelled data in an offline manner, so as to construct effective heuristics to boost a broad range of DCOP algorithms where evaluating the quality of a partial assignment is critical, such as local search or backtracking search. Furthermore, to enable decentralized model inference, we propose a distributed embedding schema of GAT-PCM where each agent exchanges only embedded vectors, and show its soundness and complexity. Finally, we demonstrate the effectiveness of our model by combining it with a local search or a backtracking search algorithm. Extensive empirical evaluations indicate that the GAT-PCM-boosted algorithms significantly outperform the state-of-the-art methods in various benchmarks. Our pretrained cost model is available at https://github.com/dyc941126/GAT-PCM.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

Colorizing a grayscale image is inherently an ill-posed problem with multi-modal uncertainty. Language-based colorization offers a natural way of interaction to reduce such uncertainty via a user-provided caption. However, the color-object coupling and mismatch issues make the mapping from word to color difficult. In this paper, we propose L-CoDe, a Language-based Colorization network using color-object Decoupled conditions. A predictor for object-color corresponding matrix (OCCM) and a novel attention transfer module (ATM) are introduced to solve the color-object coupling problem. To deal with color-object mismatch that results in incorrect color-object correspondence, we adopt a soft-gated injection module (SIM). We further present a new dataset containing annotated color-object pairs to provide supervisory signals for resolving the coupling problem. Experimental results show that our approach outperforms state-of-the-art methods conditioned on captions.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

Human-Object Interaction (HOI) detection is a fundamental task in high-level human-centric scene understanding. We propose PhraseHOI, containing a HOI branch and a novel phrase branch, to leverage language prior and improve relation expression. Specifically, the phrase branch is supervised by semantic embeddings, whose ground truths are automatically converted from the original HOI annotations without extra human efforts. Meanwhile, a novel label composition method is proposed to deal with the long-tailed problem in HOI, which composites novel phrase labels by semantic neighbors. Further, to optimize the phrase branch, a loss composed of a distilling loss and a balanced triplet loss is proposed. Extensive experiments are conducted to prove the effectiveness of the proposed PhraseHOI, which achieves significant improvement over the baseline and surpasses previous state-of-the-art methods on Full and NonRare on the challenging HICO-DET benchmark.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

Multiple modalities can provide rich semantic information; and exploiting such information will normally lead to better performance compared with the single-modality counterpart. However, it is not easy to devise an effective cross-modal fusion structure due to the variations of feature dimensions and semantics, especially when the inputs even come from different sensors, as in the field of audio-visual learning. In this work, we propose SepFusion, a novel framework that can smoothly produce optimal fusion structures for visual-sound separation. The framework is composed of two components, namely the model generator and the evaluator. To construct the generator, we devise a lightweight architecture space that can adapt to different input modalities. In this way, we can easily obtain audio-visual fusion structures according to our demands. For the evaluator, we adopt the idea of neural architecture search to select superior networks effectively. This automatic process can significantly save human efforts while achieving competitive performances. Moreover, since our SepFusion provides a series of strong models, we can utilize the model family for broader applications, such as further promoting performance via model assembly, or providing suitable architectures for the separation of certain instrument classes. These potential applications further enhance the competitiveness of our approach.

poster session 5 @ red 3, poster session 12 @ red 3, poster session 5, poster session 12

We introduce a deep architecture named HoD-Net to enable high-order differentiability for deep learning. HoD-Net is based on and generalizes the complex-step finite difference (CSFD) method. While similar to classic finite difference, CSFD approaches the derivative of a function from a higher-dimension complex domain, leading to highly accurate and robust differentiation computation without numerical stability issues. This method can be coupled with backpropagation and adjoint perturbation methods for an efficient calculation of high-order derivatives. We show how this numerical scheme can be leveraged in challenging deep learning problems, such as high-order network training, deep learning-based physics simulation, and neural differential equations.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

Automated high-stake decision-making such as medical diagnosis requires models with high interpretability and reliability. As one of the interpretable and reliable models with good prediction ability, we consider Sparse High-order Interaction Model (SHIM) in this study. However, finding statistically significant high-order interactions is challenging due to the intrinsic high dimensionality of the combinatorial effects. Another problem in data-driven modeling is the effect of "cherry-picking" a.k.a. selection bias. Our main contribution is to extend the recently developed parametric programming approach for selective inference to high-order interaction models. Exhaustive search over the cherry tree (all possible interactions) can be daunting and impractical even for a small-sized problem. We introduced an efficient pruning strategy and demonstrated the computational efficiency and statistical power of the proposed method using both synthetic and real data.

poster session 6 @ blue 5, poster session 12 @ blue 5, oral session 12 @ blue 5, poster session 6, poster session 12, oral session 12

We investigate the fairness concerns of training a machine learning model using data with missing values. Even though there are a number of fairness intervention methods in the literature, most of them require a complete training set as input. In practice, data can have missing values, and data missing patterns can depend on group attributes (e.g. gender or race). Simply applying off-the-shelf fair learning algorithms to an imputed dataset may lead to an unfair model. In this paper, we first theoretically analyze different sources of discrimination risks when training with an imputed dataset. Then, we propose an integrated approach based on decision trees that does not require a separate process of imputation and learning. Instead, we train a tree with missing incorporated as attribute (MIA), which does not require explicit imputation, and we optimize a fairness-regularized objective function. We demonstrate that our approach outperforms existing fairness intervention methods applied to an imputed dataset, through several experiments on real-world datasets.

poster session 1 @ blue 3, poster session 12 @ blue 3, oral session 12 @ blue 3, poster session 1, poster session 12, oral session 12

We study the problem of online learning with human feedback in the human-in-the-loop machine translation, in which the human translators revise the machine-generated translations and then the corrected translations are used to improve the neural machine translation (NMT) system. However, previous methods require online model updating or additional translation memory networks to achieve high-quality performance, making them inflexible and inefficient in practice. In this paper, we propose a novel non-parametric online learning method without changing the model structure. This approach introduces two k-nearest-neighbor (knn) modules: one module memorizes the human feedback, which is the correct sentences provided by human translators, while the other balances the usage of the history human feedback and original NMT models adaptively. Experiments conducted on EMEA and JRC-Acquis benchmarks demonstrate that our proposed method obtains substantial improvements on translation accuracy and achieves better adaptation performance with less repeating human correction operations.

poster session 5 @ red 5, poster session 12 @ red 5, poster session 5, poster session 12

Pansharpening is a critical yet challenging low-level vision task that aims to obtain a higher-resolution image by fusing a multispectral (MS) image and a panchromatic (PAN) image. While most pansharpening methods are based on convolutional neural network (CNN) architectures with standard convolution operations, few attempts have been made with context-adaptive/dynamic convolution, which delivers impressive results on high-level vision tasks. In this paper, we propose a novel strategy to generate local-context adaptive (LCA) convolution kernels and introduce a new global harmonic (GH) bias mechanism, exploiting image local specificity as well as integrating global information, dubbed LAGConv. The proposed LAGConv can replace the standard convolution that is context-agnostic to fully perceive the particularity of each pixel for the task of remote sensing pansharpening. Furthermore, by applying the LAGConv, we provide an image fusion network architecture, which is more effective than conventional CNN-based pansharpening approaches. The superiority of the proposed method is demonstrated by extensive experiments implemented on a wide range of datasets compared with state-of-the-art pansharpening methods. Besides, more discussions testify that the proposed LAGConv outperforms recent adaptive convolution techniques for pansharpening.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

Underwater images suffer from degradation due to light scattering and absorption. It remains challenging to restore such degraded images using deep neural networks since real-world paired data is scarcely available while synthetic paired data cannot approximate real-world data perfectly. In this paper, we propose an UnSupervised Underwater Image Restoration method (USUIR) by leveraging the homology property between a raw underwater image and a re-degraded image. Specifically, USUIR first estimates three latent components of the raw underwater image, i.e., the global background light, the transmission map, and the scene radiance (the clean image). Then, a re-degraded image is generated by randomly mixing up the estimated scene radiance and the raw underwater image. We demonstrate that imposing a homology constraint between the raw underwater image and the re-degraded image is equivalent to minimize the restoration error, and hence can be used for the unsupervised restoration. Extensive experiments show that USUIR achieves promising performance in both inference time and restoration quality.

poster session 5 @ red 2, poster session 12 @ red 2, poster session 5, poster session 12

Universal domain adaptation (UniDA) aims to transfer knowledge learned from a labeled source domain to an unlabeled target domain under domain shift and category shift. Without prior category overlap information, it is challenging to simultaneously align the common categories between two domains and separate their respective private categories. Additionally, previous studies utilize the source classifier's prediction to obtain various known labels and one generic "unknown" label of target samples. However, over-reliance on learned classifier knowledge is inevitably biased to source data, ignoring the intrinsic structure of target domain. Therefore, in this paper, we propose a novel two-stage UniDA framework called MATHS based on the principle of \textbf{M}utual ne\textbf{A}rest neighbor con\textbf{T}rast and \textbf{H}ybrid prototype di\textbf{S}crimination. In the first stage, we design an efficient mutual nearest neighbor contrastive learning scheme to achieve feature alignment, which exploits the instance-level affinity relationship to uncover the intrinsic structure of two domains. We introduce a bimodality hypothesis for the maximum discriminative probability distribution to detect the possible target private samples, and present a data-based statistical approach to separate the common and private categories. In the second stage, to obtain more reliable label predictions, we propose an incremental pseudo-classifier for target data only, which is driven by the hybrid representative prototypes. A confidence-guided prototype contrastive loss is designed to optimize the category allocation uncertainty via a self-training mechanism. Extensive experiments on three benchmarks demonstrate that MATHS outperforms previous state-of-the-arts on most UniDA settings.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

When dealing with a series of imminent issues, humans can naturally concentrate on a subset of these concerning issues by prioritizing them according to their contributions to motivational indices, e.g., the probability of winning a game. This idea of concentration offers insights into reinforcement learning of sophisticated Large-scale Multi-Agent Systems (LMAS) participated by hundreds of agents. In such an LMAS, each agent receives a long series of entity observations at each step, which can overwhelm existing aggregation networks such as graph attention networks and cause inefficiency. In this paper, we propose a concentration network called ConcNet. First, ConcNet scores the observed entities considering several motivational indices, e.g., expected survival time and state value of the agents, and then ranks, prunes, and aggregates the encodings of observed entities to extract features. Second, distinct from the well-known attention mechanism, ConcNet has a unique motivational subnetwork to explicitly consider the motivational indices when scoring the observed entities. Furthermore, we present a concentration policy gradient architecture that can learn effective policies in LMAS from scratch. Extensive experiments demonstrate that the presented architecture has excellent scalability and flexibility, and significantly outperforms existing methods on LMAS benchmarks.

poster session 5 @ red 3, poster session 12 @ red 3, oral session 5 @ red 3, poster session 5, poster session 12, oral session 5

Existing portrait matting methods either require auxiliary inputs that are costly to obtain or involve multiple stages that are computationally expensive, making them less suitable for real-time applications. In this work, we present a light-weight matting objective decomposition network (MODNet) for portrait matting in real-time with a single input image. The key idea behind our efficient design is by optimizing a series of sub-objectives simultaneously via explicit constraints. In addition, MODNet includes two novel techniques for improving model efficiency and robustness. First, an Efficient Atrous Spatial Pyramid Pooling (e-ASPP) module is introduced to fuse multi-scale features for semantic estimation. Second, a self-supervised sub-objectives consistency (SOC) strategy is proposed to adapt MODNet to real-world data to address the domain shift problem common to trimap-free methods. MODNet is easy to be trained in an end-to-end manner. It is much faster than contemporaneous methods and runs at 67 frames per second on a 1080Ti GPU. Experiments show that MODNet outperforms prior trimap-free methods by a large margin on both Adobe Matting Dataset and a carefully designed photographic portrait matting (PPM-100) benchmark proposed by us. Further, MODNet achieves remarkable results on daily photos and videos.

poster session 6 @ blue 2, poster session 12 @ blue 2, poster session 6, poster session 12

Few-shot text classification has recently been promoted by the meta-learning paradigm which aims to identify target classes with knowledge transferred from source classes with sets of small tasks named episodes. Despite their success, existing works building their meta-learner based on Prototypical Networks are unsatisfactory in learning discriminative text representations between similar classes, which may lead to contradictions during label prediction. In addition, the task-level and instance-level overfitting problems in few-shot text classification caused by a few training examples are not sufficiently tackled. In this work, we propose a contrastive learning framework named ContrastNet to tackle both discriminative representation and overfitting problems in few-shot text classification. ContrastNet learns to pull closer text representations belonging to the same class and push away text representations belonging to different classes, while simultaneously introducing unsupervised contrastive regularization at both task-level and instance-level to prevent overfitting. Experiments on 8 few-shot text classification datasets show that ContrastNet outperforms the current state-of-the-art models.

poster session 3 @ red 4, poster session 12 @ red 4, poster session 3, poster session 12

Maximum Inner Product Search (MIPS) plays an important role in many applications ranging from information retrieval, recommender systems to natural language processing and machine learning. However, exhaustive MIPS is often expensive and impractical when there are a large number of candidate items. The state-of-the-art approximated MIPS is product quantization with a score-aware loss, which weighs more heavily on items with larger inner product scores. However, it is challenging to extend the score-aware loss for additive quantization due to parallel-orthogonal decomposition of residual error. Learning additive quantization with respect to this loss is important since additive quantization can achieve a lower approximation error than product quantization. To this end, we propose a quantization method called Anisotropic Additive Quantization to combine the score-aware anisotropic loss and additive quantization. To efficiently update the codebooks in this algorithm, we develop a new alternating optimization algorithm. The proposed algorithm is extensively evaluated on three real-world datasets. The experimental results show that it outperforms the state-of-the-art baselines with respect to approximate search accuracy while guaranteeing a similar retrieval efficiency.

poster session 6 @ blue 5, poster session 12 @ blue 5, poster session 6, poster session 12

We study graph-based approaches to span-based semantic role labeling. This task is difficult due to the need to enumerate all possible predicate-argument pairs and the high degree of imbalance between positive and negative samples. Based on these difficulties, high-order inference that considers interactions between multiple arguments and predicates is often deemed beneficial but has rarely been used in span-based semantic role labeling. Because even for second-order inference, there are already O(n^5) parts for a sentence of length n, and exact high-order inference is intractable. In this paper, we propose a framework consisting of two networks: a predicate-agnostic argument pruning network that reduces the number of candidate arguments to O(n), and a semantic role labeling network with an optional second-order decoder that is unfolded from an approximate inference algorithm. Our experiments show that our framework achieves significant and consistent improvement over previous approaches.

poster session 3 @ red 4, poster session 12 @ red 4, poster session 3, poster session 12

Neural ordinary differential equations (NODE) have been proposed as a continuous depth generalization to popular deep learning models such as Residual networks (ResNets). They provide parameter efficiency and automate the model selection process in deep learning models to some extent. However, they lack the much-required uncertainty modelling and robustness capabilities which are crucial for their use in several real-world applications such as autonomous driving and healthcare. We propose a novel and unique approach to model uncertainty in NODE by considering a distribution over the end-time T of the ODE solver. The proposed approach, latent time NODE (LT-NODE), treats T as a latent variable and apply Bayesian learning to obtain a posterior distribution over T from the data. In particular, we use variational inference to learn an approximate posterior and the model parameters. Prediction is done by considering the NODE representations from different samples of the posterior and can be done efficiently using a single forward pass. As T implicitly defines the depth of a NODE, posterior distribution over T would also help in model selection in NODE. We also propose, adaptive latent time NODE (ALT-NODE), which allow each data point to have a distinct posterior distribution over end-times. ALT-NODE uses amortized variational inference to learn an approximate posterior using inference networks. We demonstrate the effectiveness of the proposed approaches in modelling uncertainty and robustness through experiments on synthetic and several real-world image classification data.

poster session 1 @ blue 3, poster session 12 @ blue 3, poster session 1, poster session 12

We propose an information-theoretic bias measurement technique through a causal interpretation of spurious correlation, which is effective to identify the feature-level algorithmic bias by taking advantage of conditional mutual information. Although several bias measurement methods have been proposed and widely investigated to achieve algorithmic fairness in various tasks such as face recognition, their accuracy- or logit-based metrics are susceptible to leading to trivial prediction score adjustment rather than fundamental bias reduction. Hence, we design a novel debiasing framework against the algorithmic bias, which incorporates a bias regularization loss derived by the proposed information-theoretic bias measurement approach. In addition, we present a simple yet effective unsupervised debiasing technique based on stochastic label noise, which does not require the explicit supervision of bias information. The proposed bias measurement and debiasing approaches are validated in diverse realistic scenarios through extensive experiments on multiple standard benchmarks.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Face self-occlusions are inevitable due to the 3D nature of the human face and the loss of information in the projection process from 3D to 2D images. While recovering face self-occlusions based on 3D face reconstruction, e.g., 3D Morphable Model (3DMM) and its variants provides an effective solution, most of the existing methods show apparent limitations in expressing high-fidelity, natural, and diverse facial details. To overcome these limitations, we propose in this paper a new generative adversarial network (MvInvert) for natural face self-occlusion recovery without using paired image-texture data. We design a coarse-to-fine generator for photorealistic texture generation. A coarse texture is computed by inpainting the invisible areas in the photorealistic but incomplete texture sampled directly from the 2D image using the unrealistic but complete statistical texture from 3DMM. Then, we design a multi-view Residual-based GAN Inversion, which re-renders and refines multi-view 2D images, which are used for extracting multiple high-fidelity textures. Finally, these high-fidelity textures are fused based on their visibility maps via Poisson blending. To perform adversarial learning to assure the quality of the recovered texture, we design a discriminator consisting of two heads, i.e., one for global and local discrimination between the recovered texture and a small set of real textures in UV space, and the other for discrimination between the input image and the re-rendered 2D face images via pixel-wise, identity, and adversarial losses. Extensive experiments demonstrate that our approach outperforms the state-of-the-art methods in face self-occlusion recovery under unconstrained scenarios.

poster session 2 @ red 1, poster session 12 @ red 1, poster session 2, poster session 12

Vision-based autonomous urban driving in dense traffic is quite challenging due to the complicated urban environment and the dynamics of the driving behaviors. Widely-applied methods either heavily rely on hand-crafted rules or learn from limited human experience, which makes them hard to generalize to rare but critical scenarios. In this paper, we present a novel CAscade Deep REinforcement learning framework, CADRE, to achieve model-free vision-based autonomous urban driving. In CADRE, to derive representative latent features from raw observations, we first offline train a Co-attention Perception Module (CoPM) that leverages the co-attention mechanism to learn the inter-relationships between the visual and control information from a pre-collected driving dataset. Cascaded by the frozen CoPM, we then present an efficient distributed proximal policy optimization framework to online learn the driving policy under the guidance of particularly designed reward functions. We perform a comprehensive empirical study with the CARLA NoCrash benchmark as well as specific obstacle avoidance scenarios in autonomous urban driving tasks. The experimental results well justify the effectiveness of CADRE and its superiority over the state-of-the-art by a wide margin.

poster session 2 @ red 1, poster session 12 @ red 1, oral session 2 @ red 1, poster session 2, poster session 12, oral session 2

We present Nonparametric Approximation of Inter-Trace returns (NAIT), a Reinforcement Learning algorithm for discrete action, pixel-based environments that is both highly sample and computation efficient. NAIT is a lazy-learning approach with an update that is equivalent to episodic Monte-Carlo on episode completion, but that allows the stable incorporation of rewards while an episode is ongoing. We make use of a fixed domain-agnostic representation, simple distance based exploration and a proximity graph-based lookup to facilitate extremely fast execution. We empirically evaluate NAIT on both the 26 and 57 game variants of ATARI100k where, despite its simplicity, it achieves competitive performance in the online setting with greater than 100x speedup in wall-time.

poster session 6 @ blue 1, poster session 12 @ blue 1, oral session 6 @ blue 1, poster session 6, poster session 12, oral session 6

Dealing with real-world reinforcement learning (RL) tasks, we shall be aware that the environment may have sudden changes. We expect that a robust policy is able to handle such changes and adapt to the new environment rapidly. Context-based meta reinforcement learning aims at learning environment adaptable policies. These methods adopt a context encoder to perceive the environment on-the-fly, following which a contextual policy makes environment adaptive decisions according to the context. However, previous methods show lagged and unstable context extraction, which are hard to handle sudden changes well. This paper proposes an environment sensitive contextual policy learning (ESCP) approach, in order to improve both the sensitivity and the robustness of context encoding. ESCP has three key components: \emph{variance minimization} that forces a rapid and stable encoding of the environment context, \emph{relational matrix determinant maximization} that avoids trivial solutions by the variance minimization, and a \emph{history-truncated recurrent neural network} model that avoids old memory interference. We use a grid-world task and $5$ locomotion controlling tasks with changing parameters to empirically assess our algorithm. Experiment results show that in environments with both in-distribution and out-of-distribution parameter changes, ESCP can not only better recover the environment encoding, but also adapt more rapidly to the post-change environment ($10\times$ faster in the grid-world) while the return performance is kept or improved, compared with state-of-the-art meta RL methods.

poster session 6 @ blue 1, poster session 12 @ blue 1, oral session 12 @ blue 1, poster session 6, poster session 12, oral session 12