Papers

For better monitoring and controlling influenza, WHO has launched FluNet (recently integrated to FluMART) to provide a unified platform for participating countries to routinely collect influenza-related syndromic, epidemiological and virological data. However, the reported data were incomplete.We propose a novel surveillance system based on data from multiple sources to accurately assess the epidemic status of different countries, especially for those with missing surveillance data in some periods. The proposed method can automatically select a small set of reliable and informative indicators for assessing the underlying epidemic status and proper supporting data to train the predictive model. Our proactive selection method outperforms three other out-of-box methods (linear regression, multilayer perceptron, and long-short term memory) to make accurate predictions.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

The modeling and control of complex physical systems are essential in real-world problems. We propose a novel framework that is generally applicable to solving PDE-constrained optimal control problems by introducing surrogate models for PDE solution operators with special regularizers. The procedure of the proposed framework is divided into two phases: solution operator learning for PDE constraints (Phase 1) and searching for optimal control (Phase 2). Once the surrogate model is trained in Phase 1, the optimal control can be inferred in Phase 2 without intensive computations. Our framework can be applied to both data-driven and data-free cases. We demonstrate the successful application of our method to various optimal control problems for different control variables with diverse PDE constraints from the Poisson equation to Burgers' equation.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

Recently, various image-to-image translation (I2I) methods have improved mode diversity and visual quality in terms of neural networks or regularization terms. However, conventional I2I methods relies on a static decision boundary and the encoded representations in those methods are entangled with each other, so they often face with ‘mode collapse’ phenomenon. To mitigate mode collapse, 1) we design a so-called style-guided discriminator that guides an input image to the target image style based on the strategy of flexible decision boundary. 2) Also, we make the encoded representations include independent domain attributes. Based on two ideas, this paper proposes Style-Guided and Disentangled Representation for Robust Image-to-Image Translation (SRIT). SRIT showed outstanding FID by 8%, 22.8%, and 10.1% for CelebA-HQ, AFHQ, and Yosemite datasets, respectively. The translated images of SRIT reflect the styles of target domain successfully. This indicates that SRIT shows better mode diversity than previous works.

poster session 6 @ red 2, poster session 7 @ red 2, poster session 6, poster session 7

We introduce an inductive logic programming approach that combines classical divide-and-conquer search with modern constraint-driven search. Our anytime approach can learn optimal, recursive, and large programs and supports predicate invention. Our experiments on three domains (classification, inductive general game playing, and program synthesis) show that our approach can increase predictive accuracies and reduce learning times.

poster session 6 @ blue 3, poster session 7 @ blue 3, oral session 7 @ blue 3, poster session 6, poster session 7, oral session 7

Developing Named Entity Recognition (NER) systems for Indian languages has been a long-standing challenge, mainly owing to the requirement of a large amount of annotated clean training instances. This paper proposes an end-to-end framework for NER for Indian languages in a low-resource setting by exploiting parallel corpora of English and Indian languages and an English NER dataset. The proposed framework includes an annotation projection method that combines word alignment score and NER tag prediction confidence score on source language (English) data to generate weakly labeled data in a target Indian language. We employ a variant of the Teacher-Student model and optimize it jointly on the pseudo labels of the Teacher model and predictions on the generated weakly labeled data. We also present manually annotated test sets for three Indian languages: Hindi, Bengali, and Gujarati. We evaluate the performance of the proposed framework on the test sets of the three Indian languages. Empirical results show a minimum 10% performance improvement compared to the zero-shot transfer learning model on all languages. This indicates that weakly labeled data generated using the proposed annotation projection method in target Indian languages can complement well-annotated source language data to enhance performance. Our code is publicly available at https://github.com/aksh555/CL-NERIL.

poster session 6 @ red 5, poster session 9 @ red 5, poster session 6, poster session 9

Advantage Learning (AL) seeks to increase the action gap between the optimal action and its competitors, so as to improve the robustness to estimation errors. However, the method becomes problematic when the optimal action induced by the approximated value function does not agree with the true optimal action. In this paper, we present a novel method, named clipped Advantage Learning (clipped AL), to address this issue. The method is inspired by our observation that increasing the action gap blindly for all given samples while not taking their necessities into account could accumulate more errors in the performance loss bound, leading to a slow value convergence, and to avoid that, we should adjust the advantage value adaptively. We show that our simple clipped AL operator not only enjoys fast convergence guarantee but also retains proper action gaps, hence achieving a good balance between the large action gap and the fast convergence. The feasibility and effectiveness of the proposed method are verified empirically on several RL benchmarks with promising performance.

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

Meta reinforcement learning (RL) attempts to discover new RL algorithms automatically from environment interaction. In so-called black-box approaches, the policy and the learning algorithm are jointly represented by a single neural network. These methods are very flexible, but they tend to underperform in terms of generalisation to new, unseen environments. In this paper, we explore the role of symmetries in meta-generalisation. We show that a recent successful meta RL approach that meta-learns an objective for backpropagation-based learning exhibits certain symmetries (specifically the reuse of the learning rule, and invariance to input and output permutations) that are not present in typical black-box meta RL systems. We hypothesise that these symmetries can play an important role in meta-generalisation. Building off recent work in black-box supervised meta learning, we develop a black-box meta RL system that exhibits these same symmetries. We show through careful experimentation that incorporating these symmetries can lead to algorithms with a greater ability to generalise to unseen action & observation spaces, tasks, and environments.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

In few-shot imitation learning (FSIL), using behavioral cloning (BC) to solve unseen tasks with few expert demonstrations becomes a popular research direction. The following capabilities are essential in robotics applications: (1) Behaving in compound tasks that contain multiple stages. (2) Retrieving knowledge from few length-variant and misalignment demonstrations. (3) Learning from a different expert. No previous work can achieve these abilities at the same time. In this work, we conduct FSIL problem under the union of above settings and introduce a novel stage conscious attention network (SCAN) to retrieve knowledge from few demonstrations simultaneously. SCAN uses an attention module to identify each stage in length-variant demonstrations. Moreover, it is designed under demonstration-conditioned policy that learns the relationship between experts and agents. Experiment results show that SCAN can learn from different experts without fine-tuning and outperform baselines in complicated compound tasks with explainable visualization.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

CNN-LSTM based architectures have played an important role in image captioning, but limited by the training efficiency and expression ability, researchers began to explore the CNN-Transformer based models and achieved great success. Meanwhile, almost all recent works adopt Faster R-CNN as the backbone encoder to extract region-level features from given images. However, Faster R-CNN needs a pre-training on an additional dataset, which divides the image captioning task into two stages and limits its potential applications. In this paper, we build a pure Transformer-based model, which integrates image captioning into one stage and realizes end-to-end training. Firstly, we adopt SwinTransformer to replace Faster R-CNN as the backbone encoder to extract grid-level features from given images; Then, referring to Transformer, we build a refining encoder and a decoder. The refining encoder refines the grid features by capturing the intra-relationship between them, and the decoder decodes the refined features into captions word by word. Furthermore, in order to increase the interaction between multi-modal (vision and language) features to enhance the modeling capability, we calculate the mean pooling of grid features as the global feature, then introduce it into refining encoder to refine with grid features together, and add a pre-fusion process of refined global feature and generated words in decoder. To validate the effectiveness of our proposed model, we conduct experiments on MSCOCO dataset. The experimental results compared to existing published works demonstrate that our model achieves new state-of-the-art performances of 138.2% (single model) and 141.0% (ensemble of 4 models) CIDEr scores on 'Karpathy' offline test split and 136.0% (c5) and 138.3% (c40) CIDEr scores on the official online test server. Trained models and source code will be released.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

Cross-Lingual Information Retrieval (CLIR) aims to rank the documents written in a language different from the user’s query. The intrinsic gap between different languages is an essential challenge for CLIR. In this paper, we introduce the multilingual knowledge graph (KG) to the CLIR task due to the sufficient information of entities in multiple languages. It is regarded as a “silver bullet” to simultaneously perform explicit alignment between queries and documents and also broaden the representations of queries. And we propose a model named CLIR with HIerarchical Knowledge Enhancement (HIKE) for our task. The proposed model encodes the textual information in queries, documents and the KG with multilingual BERT, and incorporates the KG information in the query-document matching process with a hierarchical information fusion mechanism. Particularly, HIKE first integrates the entities and their neighborhood in KG into query representations with a knowledge-level fusion, then combines the knowledge from both source and target languages to further mitigate the linguistic gap with a language-level fusion. Finally, experimental results demonstrate that HIKE achieves substantial improvements over state-of-the-art competitors.

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

We use integer programming (IP) and constraint programming (CP) to search for sets of mutually orthogonal latin squares (MOLS). We improve the performance of the solvers by formulating an extended symmetry breaking method and provide an alternative CP encoding which performs much better in practice. Using state-of-the-art solvers we are able to quickly find pairs of MOLS (or prove their nonexistence) in all orders up to and including eleven. We also analyze the effectiveness of using CP and IP solvers to search for triples of MOLS and estimate the running time of using this approach to resolve the longstanding open problem of determining the existence of a triple of MOLS of order ten.

poster session 6 @ blue 4, poster session 10 @ blue 4, poster session 6, poster session 10

Pose-invariant face recognition is a practically useful but challenging task. This paper introduces a novel method to learn pose-invariant feature representation without normalizing profile faces to frontal ones or learning disentangled features. We first design a novel strategy to learn pose-invariant feature embeddings by distilling the angular knowledge of frontal faces extracted by teacher network to student network, which enables the handling of faces with large pose variations. In this way, the features of faces across variant poses can cluster compactly for the same person to create a pose-invariant face representation. Secondly, we propose a Pose-Adaptive Angular Distillation loss to mitigate the negative effect of uneven distribution of face poses in the training dataset to pay more attention to the samples with large pose variations. Extensive experiments on two challenging benchmarks (IJB-A and CFP-FP) show that our approach consistently outperforms the existing methods.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

Large-scale pre-training has recently revolutionized vision-and-language (VL) research. Models such as LXMERT and UNITER have significantly lifted the state of the art over a wide range of VL tasks. However, the large number of parameters in such models hinders their application in practice. In parallel, work on the lottery ticket hypothesis (LTH) has shown that deep neural networks contain small matching subnetworks that can achieve on par or even better performance than the dense networks when trained in isolation. In this work, we perform the first empirical study to assess whether such trainable subnetworks also exist in pre-trained VL models. We use UNITER as the main testbed (also test on LXMERT and ViLT), and consolidate 7 representative VL tasks for experiments, including visual question answering, visual commonsense reasoning, visual entailment, referring expression comprehension, image-text retrieval, GQA, and NLVR2. Through comprehensive analysis, we summarize our main findings as follows. (i) It is difficult to find subnetworks that strictly match the performance of the full model. However, we can find relaxed winning tickets at 50%-70% sparsity that maintain 99% of the full accuracy. (ii) Subnetworks found by task-specific pruning transfer reasonably well to the other tasks, while those found on the pre-training tasks at 60%/70% sparsity transfer universally, matching 98%/96% of the full accuracy on average over all the tasks. (iii) Besides UNITER, other models such as LXMERT and ViLT can also play lottery tickets. However, the highest sparsity we can achieve for ViLT is far lower than LXMERT and UNITER (30% vs. 70%). (iv) LTH also remains relevant when using other training methods (e.g., adversarial training).

poster session 6 @ red 4, poster session 7 @ red 4, oral session 7 @ red 4, poster session 6, poster session 7, oral session 7

We consider a sequential blocked matching (SBM) model where strategic agents repeatedly report ordinal preferences over a set of services to a central planner. The planner’s goal is to elicit agents’ true preferences and design a policy that matches services to agents in order to maximize the expected social welfare with the added constraint that each matched service can be blocked or unavailable for a number of time periods. Naturally, SBM models the repeated allocation of reusable services to a set of agents where each allocated service becomes unavailable for a fixed duration We first consider the offline SBM setting, where the strategic agents are aware of their true preferences. We measure the performance of any policy by distortion, the worst-case multiplicative approximation guaranteed by any policy. For the setting with s services, we establish lower bounds of Ω(s) and Ω(√s) on the distortions of any deterministic and randomised mechanisms, respectively. We complement these results by providing approximately truthful, measured by incentive ratio, deterministic and randomised policies based on random serial dictatorship which match our lower bounds. Our results show that there is a significant improvement if one considers the class of randomised policies. Finally, we consider the online SBM setting with bandit feedback where each agent is initially unaware of her true preferences, and the planner must facilitate each agent in the learning of their preferences through the matching of services over time. We design an approximately truthful mechanism based on the Explore-then-Commit paradigm, which achieves logarithmic dynamic approximate regret.

poster session 6 @ blue 6, poster session 10 @ blue 6, oral session 6 @ blue 6, poster session 6, poster session 10, oral session 6

Graph neural networks are a promising architecture for learning and inference with graph-structured data. Yet, how to generate informative, fixed dimensional features for graphs with varying size and topology can still be challenging. Typically, this is achieved through graph-pooling, which summarizes a graph by compressing all its nodes into a single vector. Is such a “collapsing-style” graph-pooling the only choice for graph classification? From complex system’s point of view, properties of a complex system arise largely from the interaction among its components. Therefore, we speculate that preserving the interacting relation between parts, instead of pooling them together, could benefit system level prediction. To verify this, we propose SLIM, a graph neural network model for Structural Landmarking and Interaction Modelling. The main idea is to compute a set of end-to-end optimizable sub-structure landmarks, so that any input graph can be projected onto these (spatially) local structural representatives for a faithful, global characterization. By doing so, explicit interaction between component parts of a graph can be leveraged directly in generating discriminative graph representation. Encouraging results are observed on benchmark datasets for graph classification, demonstrating the value of interaction modelling in the design of graph neural networks.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

Inferring spatial relations in natural language is a crucial ability an intelligent system should possess. The bAbI dataset tries to capture tasks relevant to this domain (task 17 and 19). However, these tasks have several limitations. Most importantly, they are limited to fixed expressions, they are limited in the number of reasoning steps required to solve them, and they fail to test the robustness of models to input that contains irrelevant or redundant information. In this paper, we present a new Question-Answering dataset called StepGame for robust multi-step spatial reasoning in texts. Our experiments demonstrate that state-of-the-art models on the bAbI dataset struggle on the StepGame dataset. Moreover, we propose a Tensor-Product based Memory-Augmented Neural Network (TP-MANN) specialized for spatial reasoning tasks. Experimental results on both datasets show that our model outperforms all the baselines with superior generalization and robustness performance.

poster session 6 @ red 5, poster session 9 @ red 5, oral session 9 @ red 5, poster session 6, poster session 9, oral session 9

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

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

Answer Set Programming (ASP) is a framework in artificial intelligence and knowledge representation for declarative modeling and problem solving. Modern ASP solvers focus on the computation or enumeration of answer sets. However, a variety of probabilistic applications in reasoning or logic programming require counting answer sets. While counting can be done by enumeration, simple enumeration becomes immediately infeasible if the number of solutions is high. On the other hand, approaches to exact counting are of high worst-case complexity. In fact, in propositional model counting, exact counting becomes impractical. In this work, we present a scalable approach to approximate counting for answer set programming. Our approach is based on systematically adding XOR constraints to ASP programs, which divide the search space. We prove that adding random XOR constraints partitions the answer sets of an ASP program. In practice, we use a Gaussian elimination-based approach by lifting ideas from SAT to ASP and integrating it into a state of the art ASP solver, which we call ApproxASP. Finally, our experimental evaluation shows the scalability of our approach over the existing ASP systems.

poster session 6 @ red 2, poster session 7 @ red 2, poster session 6, poster session 7

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

In safety-critical machine learning applications, it is crucial to defend models against adversarial attacks --- small modifications of the input that change the predictions. Besides rigorously studied $\ell_p$-bounded additive perturbations, semantic perturbations (e.g. rotation, translation) raise a serious concern on deploying ML systems in real-world. Therefore, it is important to provide provable guarantees for deep learning models against semantically meaningful input transformations. In this paper, we propose a new universal probabilistic certification approach based on Chernoff-Cramer bounds that can be used in general attack settings. We estimate the probability of a model to fail if the attack is sampled from a certain distribution. Our theoretical findings are supported by experimental results on different datasets.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

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

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

Existing scene graph generation methods suffer the limitations when the image lacks of sufficient visual contexts. To address this limitation, we propose a knowledge-enhanced scene graph generation model with multimodal relation alignment, which supplements the missing visual contexts by well-aligned textual knowledge. First, we represent the textual information into contextualized knowledge which is guided by the visual objects to enhance the contexts. Furthermore, we align the multimodal relation triplets by co-attention module for better semantics fusion. The experimental results show the effectiveness of our method.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

In recent years, joint text-image embeddings have significantly improved thanks to the development of transformer-based Vision-Language models. Despite these advances, we still need to better understand the representations produced by those models. In this paper, we compare pre-trained and fine-tuned representations at a vision, language and multimodal level. To that end, we use a set of probing tasks to evaluate the performance of state-of-the-art Vision-Language models and introduce new datasets specifically for multimodal probing. These datasets are carefully designed to address a range of multimodal capabilities while minimizing the potential for models to rely on bias. Although the results confirm the ability of Vision-Language models to understand color at a multimodal level, the models seem to prefer relying on bias in text data for object position and size. On semantically adversarial examples, we find that those models are able to pinpoint fine-grained multimodal differences. Finally, we also notice that fine-tuning a Vision-Language model on multimodal tasks does not necessarily improve its multimodal ability. We make all datasets and code available to replicate experiments.

poster session 6 @ red 4, poster session 7 @ red 4, oral session 6 @ red 4, poster session 6, poster session 7, oral session 6

Black-box attacks in deep reinforcement learning usually retrain substitute policies to mimic behaviors of target policies as well as craft adversarial examples, and attack the target policies with these transferable adversarial examples. However, the transferability of adversarial examples is not always guaranteed. Moreover, current methods of crafting adversarial examples only utilize simple pixel space metrics which neglect semantics in the whole images, and thus generate unnatural adversarial examples. To address these problems, we propose an advRL-GAN framework to directly generate semantically natural adversarial examples in the black-box setting, bypassing the transferability requirement of adversarial examples. It formalizes the black-box attack as a reinforcement learning (RL) agent, which explores natural and aggressive adversarial examples with generative adversarial networks and the feedback of target agents. To the best of our knowledge, it is the first RL-based adversarial attack on a deep RL agent. Experimental results on multiple environments demonstrate the effectiveness of advRL-GAN in terms of reward reductions and magnitudes of perturbations, and validate the sparse and targeted property of adversarial perturbations through visualization.

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

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

The use of priors to avoid manual labeling for training machine learning methods has received much attention in the last few years. One of the critical subthemes in this regard is Learning from Label Proportions (LLP), where only the information about class proportions is available for training the models. While various LLP training settings verse in the literature, most approaches focus on bag-level label proportions errors, often leading to suboptimal solutions. This paper proposes a new model that jointly uses prototypical contrastive learning and bag-level cluster proportions to implement efficient LLP classification. Our proposal explicitly relaxes the equipartition constraint commonly used in prototypical contrastive learning methods and incorporates the exact cluster proportions into the optimal transport algorithm used for cluster assignments. At inference time, we compute the clusters' assignment, delivering instance-level classification. We experimented with our method on two widely used image classification benchmarks and report a new state-of-art LLP performance, achieving results close to fully supervised methods.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

Motion completion, as a challenging and fundamental problem, is of great significance in film and game applications. For different motion completion application scenarios (in-betweening, in-filling, and blending), most previous methods deal with the completion problems with case-by-case methodology designs. In this work, we propose a simple but effective method to solve multiple motion completion problems under a unified framework and achieves a new state-of-the-art accuracy on LaFAN1 (+17% better than previous sota) under multiple evaluation settings. Inspired by the recent great success of self-attention-based transformer models, we consider the completion as a sequence-to-sequence prediction problem. Our method consists of three modules - a standard transformer encoder with self-attention that learns long-range dependencies of input motions, a trainable mixture embedding module that models temporal information and encodes different key-frame combinations in a unified form, and a new motion perceptual loss for better capturing high-frequency movements. Our method can predict multiple missing frames within a single forward propagation in real-time and get rid of the post-processing requirement. We also introduce a novel large-scale dance movement dataset for exploring the scaling capability of our method and its effectiveness in complex motion applications.

poster session 6 @ red 6, poster session 11 @ red 6, oral session 11 @ red 6, poster session 6, poster session 11, oral session 11

To synthesize images with preferred objects and interactions, a controllable way is to generate the image from a scene graph and a large pool of object crops, where the spatial arrangements of the objects in the image are defined by the scene graph while the object appearances are determined by the retrieved crops from the pool. In this paper, we propose a novel framework with such a semi-parametric generation strategy. First, to encourage the retrieval of mutually compatible crops, we design a sequential selection strategy where the crop selection for each object is determined by the contents and locations of all object crops that have been chosen previously. Such process is implemented via a transformer trained with contrastive losses. Second, to generate the final image, our hierarchical generation strategy leverages hierarchical gated convolutions which are employed to synthesize areas not covered by any image crops, and a patch-guided spatially adaptive normalization module (PSANM) which is proposed to guarantee the final generated images complying with the crop appearance and the scene graph. Evaluated on the challenging Visual Genome and COCO-Stuff dataset, our experimental results demonstrate the superiority of our proposed method over existing state-of-the-art methods.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

Reinforcement Learning (RL) enables agents to learn how to perform various tasks from scratch. In domains like autonomous driving, recommendation systems, and more, optimal RL policies learned could cause a privacy breach if the policies memorize any part of the private reward. We study the set of existing differentially-private RL policies derived from various RL algorithms such as Value Iteration, Deep-Q Networks, and Vanilla Proximal Policy Optimization. We propose a new Privacy-Aware Inverse RL analysis framework (PRIL) that involves performing reward reconstruction as an adversarial attack on private policies that the agents may deploy. For this, we introduce the reward reconstruction attack, wherein we seek to reconstruct the original reward from a privacy-preserving policy using the Inverse RL algorithm. An adversary must do poorly at reconstructing the original reward function if the agent uses a tightly private policy. Using this framework, we empirically test the effectiveness of the privacy guarantee offered by the private algorithms on instances of the FrozenLake domain of varying complexities. Based on the analysis performed, we infer a gap between the current standard of privacy offered and the standard of privacy needed to protect reward functions in RL. We do so by quantifying the extent to which each private policy protects the reward function by measuring distances between the original and reconstructed rewards.

poster session 6 @ red 1, poster session 10 @ red 1, oral session 10 @ red 1, poster session 6, poster session 10, oral session 10

The simple temporal problem (STP) is one of the most influential reasoning formalisms for representing temporal information in AI. We study the problem of resolving inconsistency of data encoded in the STP. We prove that the problem of identifying a maximally large consistent subset of data is NP-hard. In practical instances, it is reasonable to assume that the amount of erroneous data is small. We therefore parameterize by the number of constraints that need to be removed to achieve consistency. Using tools from parameterized complexity we design fixed-parameter tractable algorithms for two large fragments of the STP. Our main algorithmic results employ reductions to the Directed Subset Feedback Arc Set problem and iterative compression combined with an efficient algorithm for the Edge Multicut problem. We complement our algorithmic results with hardness results that rule out fixed-parameter tractable algorithms for all remaining non-trivial fragments of the STP (under standard complexity-theoretic assumptions). Together, our results give a full classification of the classical and parameterized complexity of the problem.

poster session 6 @ blue 4, poster session 10 @ blue 4, poster session 6, poster session 10

Graph Neural Networks (GNNs) are the first choice methods for graph machine learning problems thanks to their ability to learn state-of-the-art level representations from graph-structured data. However, centralizing a massive amount of real-world graph data for GNN training is prohibitive due to user-side privacy concerns, regulation restrictions, and commercial competition. Federated Learning is the de-facto standard for collaborative training of machine learning models over many distributed edge devices without the need for centralization. Nevertheless, training graph neural networks in a federated setting is vaguely defined and brings statistical and systems challenges. This work proposes SpreadGNN, a novel multi-task federated training framework capable of operating in the presence of partial labels and absence of a central server for the first time in the literature. We provide convergence guarantees and empirically demonstrate the efficacy of our framework on a variety of non-I.I.D. distributed graph-level molecular property prediction datasets with partial labels. Our results show that SpreadGNN outperforms GNN models trained over a central server-dependent federated learning system, even in constrained topologies.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

The task of audio-visual sound source localization has been well studied under constrained scenes, where the audio recordings are clean. However, in real-world scenarios, audios are usually contaminated by off-screen sound and background noise. They will interfere with the procedure of identifying desired sources and building visual-sound connections, making previous studies non-applicable. In this work, we propose the Interference Eraser (IEr) framework, which tackles the problem of audio-visual sound source localization in the wild. The key idea is to eliminate the interference by redefining and carving discriminative audio representations. Specifically, we observe that the previous practice of learning only a single audio representation is insufficient due to the additive nature of audio signals. We thus extend the audio representation with our Audio-Instance-Identifier module, which clearly distinguishes sounding instances when audio signals of different volumes are unevenly mixed. Then we erase the influence of the audible but off-screen sounds and the silent but visible objects by a Cross-modal Referrer module with cross-modality distillation. Quantitative and qualitative evaluations demonstrate that our proposed framework achieves superior results on sound localization tasks, especially under real-world scenarios. Code is available at https://github.com/alvinliu0/Visual-Sound-Localization-in-the-Wild.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

Reward machines are an established tool for dealing with reinforcement learning problems in which rewards are sparse and depend on complex sequences of actions. However, existing algorithms for learning reward machines assume an overly idealized setting where rewards have to be free of noise. To overcome this practical limitation, we introduce a novel type of reward machines, called stochastic reward machines, and an algorithm for learning them. Our algorithm, based on constraint solving, learns minimal stochastic reward machines from the explorations of a reinforcement learning agent. This algorithm can easily be paired with existing reinforcement learning algorithms for reward machines and guarantees to converge to an optimal policy in the limit. We demonstrate the effectiveness of our algorithm in two case studies and show that it outperforms both existing methods and a naive approach for handling noisy reward functions.

poster session 6 @ red 2, poster session 7 @ red 2, poster session 6, poster session 7

Despite the great potential of reinforcement learning (RL) in solving complex decision-making problems, generalization remains one of its key challenges, leading to difficulty in deploying learned RL policies to new environments. In this paper, we propose to improve the generalization of RL algorithms through fusing Self-supervised learning into Intrinsic Motivation (SIM). Specifically, SIM boosts representation learning through driving the cross-correlation matrix between the embeddings of augmented and non-augmented samples close to the identity matrix. This aims to increase the similarity between the embedding vectors of a sample and its augmented version while minimizing the redundancy between the components of these vectors. Meanwhile, the redundancy reduction based self-supervised loss is converted to an intrinsic reward to further improve generalization in RL via an auxiliary objective. As a general paradigm, SIM can be implemented on top of any RL algorithm. Extensive evaluations have been performed on a diversity of tasks. Experimental results demonstrate that SIM consistently outperforms the state-of-the-art methods and exhibits superior generalization capability and sample efficiency.

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

The paper proposes a probabilistic generalization of the well-known Strong Backdoor Set~(SBS) concept applied to the Boolean Satisfiability Problem (SAT). We call a set of Boolean variables $B$ a $\rho$-backdoor, if for a fraction $\rho$ of possible assignments of variables from $B$, assigning their values to variables in a Boolean formula in Conjunctive Normal Form (CNF) results in polynomially solvable formulas. Clearly, a $\rho$-backdoor with $\rho=1$ is an SBS. For a given set $B$ it is possible to efficiently construct an $(\varepsilon, \delta)$-approximation of parameter $\rho$ using the Monte Carlo method. Thus, we define an $(\varepsilon, \delta)$-SBS as such a set $B$ for which the estimation of the parameter $\rho$ deviates from 1 by no more than $\varepsilon$ with probability no less than $1-\delta$. We consider the problems of finding the minimum SBS and the minimum $(\varepsilon, \delta)$-SBS. To solve the former, one can use the algorithm described by R.~Williams, C.~Gomes and B.~Selman in 2003. In the paper we propose a new probabilistic algorithm to solve the latter, and show that the asymptotic estimation of the worst case complexity of the proposed algorithm is significantly smaller than that of the algorithm by Williams~et~al. For practical applications, we suggest a metaheuristic optimization algorithm based on the penalty function method to seek the minimal $(\varepsilon, \delta)$-SBS. Results of computational experiments show that the use of $(\varepsilon, \delta)$-SBSes found by the proposed algorithm allows speeding up solving of test problems related to equivalence checking and hard crafted and combinatorial benchmarks compared to state-of-the-art SAT solvers.

poster session 6 @ blue 4, poster session 10 @ blue 4, oral session 6 @ blue 4, poster session 6, poster session 10, oral session 6

We explore deep clustering of multilingual text representations for unsupervised model interpretation and induction of syntax. As these representations are high-dimensional, out-of-the-box methods like K-means do not work well. Thus, our approach jointly transforms the representations into a lower-dimensional cluster-friendly space and clusters them. We consider two notions of syntax: Part of Speech Induction (POSI) and Constituency Labelling (CoLab) in this work. Interestingly, we find that Multilingual BERT (mBERT) contains surprising amount of syntactic knowledge of English; possibly even as much as English BERT (E-BERT). Our model can be used as a supervision-free probe which is arguably a less-biased way of probing. We find that unsupervised probes show benefits from higher layers as compared to supervised probes. We further note that our unsupervised probe utilizes E-BERT and mBERT representations differently, especially for POSI. We validate the efficacy of our probe by demonstrating its capabilities as a unsupervised syntax induction technique. Our probe works well for both syntactic formalisms by simply adapting the input representations. We report competitive performance of our probe on 45-tag English POSI, state-of-the-art performance on 12-tag POSI across 10 languages, and competitive results on CoLab. We also perform zero-shot syntax induction on resource impoverished languages and report strong results.

poster session 6 @ red 5, poster session 9 @ red 5, poster session 6, poster session 9

We propose a novel scene flow estimation approach to capture and infer 3D motions from point clouds. Estimating 3D motions for point clouds is challenging, since a point cloud is unordered and its density is significantly non-uniform. Such unstructured data poses difficulties in matching corresponding points between point clouds, leading to inaccurate flow estimation. We propose a novel architecture named Sparse Convolution-Transformer Network (SCTN) that equips the sparse convolution with the transformer. Specifically, by leveraging the sparse convolution, SCTN transfers irregular point cloud into locally consistent flow features for estimating spatially consistent motions within an object/local object part. We further propose to explicitly learn point relations using a point transformer module, different from exiting methods. We show that the learned relation-based contextual information is rich and helpful for matching corresponding points, benefiting scene flow estimation. In addition, a novel loss function is proposed to adaptively encourage flow consistency according to feature similarity. Extensive experiments demonstrate that our proposed approach achieves a new state of the art in scene flow estimation. Our approach achieves an error of 0.038 and 0.037 (EPE3D) on FlyingThings3D and KITTI Scene Flow respectively, which significantly outperforms previous methods by large margins.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

Information diffusion in social networks is a well-studied concept in social choice theory. We propose the study of the diffusion of two secrets in a heterogeneous environment from the complexity perspective, that is, there are two different networks with the same set of agents (e.g., the structure of the set of followers might be different in two distinct social networks). Formally, our model combines two group identification processes for which we do have independent desiderata---either constructive, where we would like a given group of agents to be exposed to a secret, or destructive, where a given group of agents should not be exposed to a secret. To be able to reach these targets, we can either delete an agent or introduce a previously latent agent. Our results are mostly negative---all of the problems are NP-hard. Therefore, we propose a parameterized study with respect to the natural parameters, the number of influenced agents, the size of the required/protected agent sets, and the duration of the diffusion process. Most of the studied problems remain W[1]-hard even for a combination of these parameters. We complement these results with nearly optimal XP algorithms.

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

A major obstacle to achieving global convergence in distributed and federated learning is the misalignment of gradients across clients, or mini-batches due to heterogeneity and stochasticity of the distributed data. In this work, we show that data heterogeniety can in fact be exploited to improve generalization performance through implicit regularization. One way to alleviate the effects of heterogeniety is to encourage the alignment of gradients across different clients throughout training. Our analysis reveals that this goal can be accomplished by utilizing the right optimization method that replicates the implicit regularization effect of SGD, leading to gradient alignment as well as improvements in test accuracies. Since the existence of this regularization in SGD completely relies on the sequential use of different mini-batches during training, it is inherently absent when training with large mini-batches. To obtain the generalization benefits of this regularization while increasing parallelism, we propose a novel GradAlign algorithm that induces the same implicit regularization while allowing the use of arbitrarily large batches in each update. We experimentally validate the benefits of our algorithm in different distributed and federated learning settings.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

In multiwinner approval voting, the goal is to select k-member committees based on voters' approval ballots. A well-studied concept of proportionality in this context is the justified representation (JR) axiom, which demands that no large cohesive group of voters remains unrepresented. However, the JR axiom may conflict with other desiderata, such as coverage (maximizing the number of voters who approve at least one committee member) or social welfare (maximizing the number of approvals obtained by committee members). In this work, we investigate the impact of imposing the JR axiom (as well as the more demanding EJR axiom) on social welfare and coverage. Our approach is threefold: we derive worst-case bounds on the loss of welfare/coverage that is caused by imposing JR, study the computational complexity of finding 'good' committees that provide JR (obtaining a hardness result, an approximation algorithm, and an exact algorithm for one-dimensional preferences), and examine this setting empirically on several synthetic datasets.

poster session 6 @ blue 6, poster session 10 @ blue 6, poster session 6, poster session 10

The majority of popular graph kernels is based on the concept of Haussler's R-convolution kernel and defines graph similarities in terms of mutual substructures. In this work, we enrich these similarity measures by considering graph filtrations: Using meaningful orders on the set of edges, which allow to construct a sequence of nested graphs, we can consider a graph at multiple granularities. For one thing, this provides access to features on different levels of resolution. Furthermore, rather than to simply compare frequencies of features in graphs, it allows for their comparison in terms of when and for how long they exist in the sequences. In this work, we propose a family of graph kernels that incorporate these existence intervals of features. While our approach can be applied to arbitrary graph features, we particularly highlight Weisfeiler-Lehman vertex labels, leading to efficient kernels. We show that using Weisfeiler-Lehman labels over certain filtrations strictly increases the expressive power over the ordinary Weisfeiler-Lehman procedure in terms of deciding graph isomorphism. In fact, this result directly yields more powerful graph kernels based on such features and has implications to graph neural networks due to their close relationship to the Weisfeiler-Lehman method. We empirically validate the expressive power of our graph kernels and show significant improvements over state-of-the-art graph kernels in terms of predictive performance on various real-world benchmark datasets.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

Phrase grounding is a multi-modal problem that localizes a particular noun phrase in an image referred to by a text query. In the challenging zero-shot phrase grounding setting, the existing state-of-the-art grounding models have limited capacity in handling the unseen phrases. Humans, however, can ground novel types of objects in images with little effort, significantly benefiting from reasoning with commonsense. In this paper, we design a novel phrase grounding architecture that builds multi-modal knowledge graphs using external knowledge and then performs graph reasoning and spatial relation reasoning to localize the referred nouns phrases. We perform extensive experiments on different zero-shot grounding splits sub-sampled from the Flickr30K Entity and Visual Genome dataset, demonstrating that the proposed framework is orthogonal to backbone image encoders and outperforms the baselines by 2$\sim$3\% in accuracy, resulting in a significant improvement under the standard evaluation metrics.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

Tensor factorization and distanced based models play important roles in knowledge graph completion (KGC). However, the relational matrices in KGC methods often induce a high model complexity, bearing a high risk of overfitting. As a remedy, researchers propose a variety of different regularizers such as the tensor nuclear norm regularizer. Our motivation is based on the observation that the previous work only focuses on the “size” of the parametric space, while leaving the implicit semantic information widely untouched. To address this issue, we propose a new regularizer, namely, Equivariance Regularizer (ER), which can suppress overfitting by leveraging the implicit semantic information. Specifically, ER can enhance the generalization ability of the model by employing the semantic equivariance between the head and tail entities. Moreover, it is a generic solution for both distance based models and tensor factorization based models. Our experimental results indicate a clear and substantial improvement over the state-of-the-art relation prediction methods.

poster session 6 @ blue 3, poster session 7 @ blue 3, oral session 6 @ blue 3, poster session 6, poster session 7, oral session 6

Fossil energy products such as liquefied natural gas (LNG) are among Canada's most important exports. Canadian engineers devote themselves to constructing visual surveillance systems for detecting potential LNG emissions in energy facilities. Beyond the previous infrared (IR) surveillance system, in this paper, a multimodal fusion-based LNG detection (MFLNGD) framework is proposed to enhance the detection quality by the integration of IR and visible (VI) cameras. Besides, a Fourier transformer is developed to fuse IR and VI features better. The experimental results suggest the effectiveness of the proposed framework.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

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

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

Pre-training and fine-tuning have become the \textit{de facto} paradigm in many natural language processing (NLP) tasks. However, compared to other NLP tasks, neural machine translation (NMT) aims to generate target language sentences through the contextual representation from the source language counterparts. This characteristic means the optimization objective of NMT is far from that of the universal pre-trained models (PTMs), leading to the standard procedure of pre-training and fine-tuning does not work well in NMT. In this paper, we propose a novel framework to \textit{deep fuse} the pre-trained representation into NMT, fully exploring the potential of PTMs in NMT. Specifically, we directly replace the randomly initialized Transformer encoder with a pre-trained encoder and propose a layer-wise coordination structure to coordinate PTM and NMT decoder learning. Then, we introduce a partitioned multi-task learning method to fine-tune the pre-trained parameter, reducing the gap between PTM and NMT by progressively learning the task-specific representation. Experimental results show that our approach achieves considerable improvements on WMT14 En$\rightarrow$De, WMT14 En$\rightarrow$Fr, and WMT16 Ro$\rightarrow$En translation benchmarks and outperforms previous work in both autoregressive and non-autoregressive NMT models.

poster session 6 @ red 5, poster session 9 @ red 5, poster session 6, poster session 9

We propose a new problem setting to study the sequential interactions between a recommender system and a user. Instead of assuming the user is omniscient, static, and explicit, as the classical practice does, we sketch a more realistic user behavior model, under which the user: 1) rejects recommendations if they are clearly worse than others; 2) updates her utility estimation based on rewards from her accepted recommendations; 3) withholds realized rewards from the system. We formulate the interactions between the system and such an explorative user in a $K$-armed bandit framework and study the problem of learning the optimal recommendation on the system side. We show that efficient system learning is still possible but is more difficult. In particular, the system can identify the best arm with probability at least $1-\delta$ within $O(1/\delta)$ interactions, and we prove this is tight. Our finding contrasts the result for the problem of best arm identification with fixed confidence, in which the best arm can be identified with probability $1-\delta$ within $O(\log(1/\delta))$ interactions. This gap illustrates the inevitable cost the system has to pay when it learns from an explorative user's revealed preferences on its recommendations rather than from the realized rewards.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

Neural Architecture Search (NAS) algorithms automate the task of finding optimal deep learning architectures given an initial search space of possible operations. Developing these search spaces is usually a manual affair with pre-optimized search spaces being more efficient, rather than searching from scratch. In this paper we present a new framework called Neural Architecture Type System (NeuralArTS) that categorizes the infinite set of network operations in a structured type system. We further demonstrate how NeuralArTS can be applied to convolutional layers and propose several future directions.

poster session 6 @ blue 4, poster session 10 @ blue 4, poster session 6, poster session 10

Learning information-rich and generalizable representations effectively from unlabeled multivariate cardiac signals to identify abnormal heart rhythms (cardiac arrhythmias) is valuable in real-world clinical settings but often challenging due to its complex temporal dynamics. Cardiac arrhythmias can vary significantly in temporal patterns even for the same patient (i.e., intra subject difference). Meanwhile, the same type of cardiac arrhythmia can show different temporal patterns among different patients due to different cardiac structures (i.e., inter subject difference). In this paper, we address the challenges by proposing an Intra-inter Subject self-supervised Learning (ISL) model that is customized for multivariate cardiac signals. Our proposed ISL model integrates medical knowledge into self-supervision to effectively learn from intra-inter subject differences. In intra subject self-supervision, ISL model first extracts heartbeat-level features from each subject using a channel-wise attentional CNN-RNN encoder. Then a stationarity test module is employed to capture the temporal dependencies between heartbeats. In inter subject self-supervision, we design a set of data augmentations according to the clinical characteristics of cardiac signals and perform contrastive learning among subjects to learn distinctive representations for various types of patients. Extensive experiments on three real-world datasets were conducted. In a semi-supervised transfer learning scenario, our pre-trained ISL model leads about 10% improvement over supervised training when only 1% labeled data is available, suggesting strong generalizability and robustness of the model.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

The detection of fake news often requires sophisticated reasoning skills, such as logically combining information by considering word-level subtle clues. In this paper, we move towards fine-grained reasoning for fake news detection by better reflecting the logical processes of human thinking and enabling the modeling of subtle clues. In particular, we propose a fine-grained reasoning framework by following the human’s information-processing model, introduce a mutual-reinforcement-based method for incorporating human knowledge about which evidence is more important, and design a prior-aware bi-channel kernel graph network to model subtle differences between pieces of evidence. Extensive experiments show that our model outperforms the state-of-the-art methods and demonstrate the explainability of our approach.

poster session 6 @ red 6, poster session 11 @ red 6, oral session 11 @ red 6, poster session 6, poster session 11, oral session 11

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

Popular first-order stochastic optimization methods for deep neural networks (DNNs) are usually either accelerated schemes (e.g. stochastic gradient descent (SGD) with momentum) or adaptive step-size methods (e.g. Adam/AdaMax, AdaBelief). In many contexts, including image classification with DNNs, adaptive methods tend to generalize poorly compared to SGD, i.e. get stuck in non-robust local minima; however, SGD typically converges slower. We analyze possible reasons for this behavior by modeling gradient updates as vectors of random variables and comparing them to probabilistic bounds to identify "meaningful" updates. Through experiments, we observe that only layers close to the output have "definitely non-random" update behavior. In the future, the tools developed here may be useful in rigorously quantifying and analyzing intuitions about why some optimizers and particular DNN architectures perform better than others.

poster session 6 @ blue 4, poster session 10 @ blue 4, poster session 6, poster session 10

We study the verification problem for closed-loop dynamical systems with neural-network controllers (NNCS). This problem is commonly reduced to computing the set of reachable states. When considering dynamical systems and neural networks in isolation, there exist precise approaches for that task based on set representations respectively called Taylor models and zonotopes. However, the combination of these approaches to NNCS is non-trivial because, when converting between the set representations, dependency information gets lost in each control cycle and the accumulated approximation error quickly renders the result useless. We present an algorithm to chain approaches based on Taylor models and zonotopes, yielding a precise reachability algorithm for NNCS. Because the algorithm only acts at the interface of the isolated approaches, it is applicable to general dynamical systems and neural networks and can benefit from future advances in these areas. Our implementation delivers state-of-the-art performance and is the first to successfully analyze all benchmark problems of an annual reachability competition for NNCS.

poster session 6 @ blue 3, poster session 7 @ blue 3, oral session 6 @ blue 3, poster session 6, poster session 7, oral session 6

In this paper, we propose two new definitions of local differential privacy for belief functions. One is based on Shafer’s semantics of randomly coded messages and the other from the perspective of imprecise probabilities. We show that such basic properties as composition and post-processing also hold for our new definitions. Moreover, we provide a hypothesis testing framework for these definitions and study the effect of "don’t know" in the trade-off between privacy and utility in discrete distribution estimation.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

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

Control Barrier Functions have been extensively studied to ensure guaranteed safety during inter-robot interactions. In this paper, we introduce the Parametric-Control Barrier Function (Parametric-CBF), a novel variant of the traditional Control Barrier Function to extend its expressivity in describing different safe behaviors among heterogeneous robots. A parametric-CBF based framework is presented to enable the ego robot to model the neighboring robots behavior and further improve the coordination efficiency during interaction while enjoying formally provable safety guarantees. We demonstrate the usage of Parametric-CBF in behavior prediction and adaptive safe control in the ramp merging scenario.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

Existing Scene Text Recognition (STR) methods typically use a language model to optimize the joint probability of the 1D character sequence predicted by a visual recognition (VR) model, which ignore the 2D spatial context of visual semantics within and between character instances, making them not generalize well to arbitrary shape scene text. To address this issue, we make the first attempt to perform textual reasoning based on visual semantics in this paper. Technically, given the character segmentation maps predicted by a VR model, we construct a subgraph for each instance, where nodes represent the pixels in it and edges are added between nodes based on their spatial similarity. Then, these subgraphs are sequentially connected by their root nodes and merged into a complete graph. Based on this graph, we devise a graph convolutional network for textual reasoning (GTR) by supervising it with a cross-entropy loss. GTR can be easily plugged in representative STR models to improve their performance owing to better textual reasoning. Specifically, we construct our model, namely S-GTR, by paralleling GTR to the language model in a segmentation-based STR baseline, which can effectively exploit the visual-linguistic complementarity via mutual learning. S-GTR sets new state-of-the-art on six challenging STR benchmarks and generalizes well to multi-linguistic datasets. Code is available at https://github.com/adeline-cs/GTR.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

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

Discovering the underneath causal relations is the fundamental ability for reasoning about the surrounding environment and predicting the future states in the physical world. Counterfactual prediction from visual input, which requires simulating future states based on unrealized situations in the past, is a vital component in causal relation tasks. In this paper, we work on the confounders that have effect on the physical dynamics, including masses, friction coefficients, etc., to bridge relations between the intervened variable and the affected variable whose future state may be altered. We propose a neural network framework combining Global Causal Relation Attention (GCRA) and Confounder Transmission Structure (CTS). The GCRA looks for the latent causal relations between different variables and estimates the confounders by capturing both spatial and temporal information. The CTS integrates and transmits the learnt confounders in a residual way, so that the estimated confounders can be encoded into the network as a constraint for object positions when performing counterfactual prediction. Without any access to ground truth information about confounders, our model outperforms the state-of-the-art method on various benchmarks by fully utilizing the constraints of confounders. Extensive experiments demonstrate that our model can generalize to unseen environments and maintain good performance.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

With the rapid increase of multimedia data, a large body of literature has emerged to work on multimodal summarization, the majority of which target at refining salient information from textual and image modalities to output a pictorial summary with the most relevant images. Existing methods mostly focus on either extractive or abstractive summarization and rely on the presence and quality of image captions to build image references. We are the first to propose a Unified framework for Multimodal Summarization grounding on BART, UniMS, that integrates extractive and abstractive objectives, as well as selecting the image output. Specially, we adopt knowledge distillation from a vision-language pretrained model to improve image selection, which avoids any requirement on the existence and quality of image captions. Besides, we introduce a visual guided decoder to better integrate textual and visual modalities in guiding abstractive text generation. Results show that our best model achieves a new state-of-the-art result on a large-scale benchmark dataset. The newly involved extractive objective as well as the knowledge distillation technique are proven to bring a noticeable improvement to the multimodal summarization task.

poster session 6 @ red 5, poster session 9 @ red 5, poster session 6, poster session 9

Realistic fine-grained multi-agent simulation of real world complex systems is crucial for many downstream tasks such as reinforcement learning. Recent work has used generative models (GANs in particular) for providing high fidelity simulation of real world systems. However, such generative models are often monolithic and miss out on modeling the interaction in multi-agent systems. In this work, we take a first step towards building multiple interacting generative models (GANs) that reflects the interaction in real world. We build and analyze a hierarchical set-up where a higher level GAN is conditioned on the output of multiple lower level GANs. We present a technique of using feedback from the higher level GAN to improve performance of lower level GANs. We mathematically characterize the conditions under which our technique is impactful, including understanding the transfer learning nature of our set-up. We present three distinct experiments on synthetic data, time-series data, and image domain, revealing the wide applicability of our technique.

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

Recent years have witnessed a large amount of decentralized data in multiple (edge) devices of end-users, while the aggregation of the decentralized data remains difficult for machine learning jobs due to laws or regulations. Federated Learning (FL) emerges as an effective approach to handling decentralized data without sharing the sensitive raw data, while collaboratively training global machine learning models. The servers in FL need to select (and schedule) devices during the training process. However, the scheduling of devices for multiple jobs with FL remains a critical and open problem. In this paper, we propose a novel multi-job FL framework to enable the parallel training process of multiple jobs. The framework consists of a system model and two scheduling methods. In the system model, we propose a parallel training process of multiple jobs, and construct a cost model based on the training time and the data fairness of various devices during the training process of diverse jobs. We propose a reinforcement learning-based method and a Bayesian optimization-based method to schedule devices for multiple jobs while minimizing the cost. We conduct extensive experimentation with multiple jobs and datasets. The experimental results show that our proposed approaches significantly outperform baseline approaches in terms of training time (up to 8.67 times faster) and accuracy (up to 44.6% higher).

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

Referring expression comprehension aims at grounding the object in an image referred to by the expression. Scene text that serves as an identifier has a natural advantage in referring to objects. However, existing methods only consider the text in the expression, but ignore the text in the image, leading to a mismatch. In this paper, we propose a novel model that can recognize the scene text. We assign the extracted scene text to its corresponding visual region and ground the target object guided by expression. Experimental results on two benchmarks demonstrate the effectiveness of our model.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

Semantic segmentation is an important task for scene understanding in self-driving cars and robotics, which aims to assign dense labels for all pixels in the image. Existing work typically improves semantic segmentation performance by exploring different network architectures on a target dataset. Little attention has been paid to build a unified system by simultaneously learning from multiple datasets due to the inherent distribution shift across different datasets. In this paper, we propose a simple, flexible, and general method for semantic segmentation, termed Cross-Dataset Collaborative Learning (CDCL). Our goal is to train a unified model for improving the performance in each dataset by leveraging information from all the datasets. Specifically, we first introduce a family of Dataset-Aware Blocks (DAB) as the fundamental computing units of the network, which help capture homogeneous convolutional representations and heterogeneous statistics across different datasets. Second, we present a Dataset Alternation Training (DAT) mechanism to facilitate the collaborative optimization procedure. We conduct extensive evaluations on diverse semantic segmentation datasets for autonomous driving. Experiments demonstrate that our method consistently achieves notable improvements over prior single-dataset and cross-dataset training methods without introducing extra FLOPs. Particularly, with the same architecture of PSPNet (ResNet-18), our method outperforms the single-dataset baseline by 5.65\%, 6.57\%, and 5.79\% mIoU on the validation sets of Cityscapes, BDD100K, CamVid, respectively. We also apply CDCL for point cloud 3D semantic segmentation and achieve improved performance, which further validates the superiority and generality of our method. Code and models will be released.

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

Image-level weakly supervised semantic segmentation (WSSS) is a fundamental yet challenging computer vision task facilitating scene understanding and automatic driving. Most existing methods resort to classification-based Class Activation Maps (CAMs) to play as the initial pseudo labels, which tend to focus on the discriminative image regions and lack customized characteristics for the segmentation task. To alleviate this issue, we propose a novel activation modulation and recalibration (AMR) scheme, which leverages a spotlight branch and a compensation branch to obtain weighted CAMs that can provide recalibration supervision and task-specific concepts. Specifically, an attention modulation module (AMM) is employed to rearrange the distribution of feature importance from the channel-spatial sequential perspective, which helps to explicitly model channel-wise interdependencies and spatial encodings to adaptively modulate segmentation-oriented activation responses. Furthermore, we introduce a cross pseudo supervision for dual branches, which can be regarded as a semantic similar regularization to mutually refine two branches. Extensive experiments show that AMR establishes a new state-of-the-art performance on the PASCAL VOC 2012 dataset, surpassing not only current methods trained with the image-level of supervision but also some methods relying on stronger supervision, such as saliency label. Experiments also reveal that our scheme is plug-and-play and can be incorporated with other approaches to boost their performance.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

In this paper, we propose a vision-language pre-training model, Clinical-BERT, for the medical domain, and devise three domain-specific tasks: Clinical Diagnosis (CD), Masked MeSH Modeling (MMM), Image-MeSH Modeling (IMM), together with one general pre-training task: Masked Language Modeling (MLM), to pre-train the model. The CD task helps the model to learn medical domain knowledge by predicting disease from radiographs. Medical Subject Headings (MeSH) words are important semantic components in radiograph reports, and the MMM task helps the model focus on the prediction of MeSH words. The IMM task helps the model learn the alignment of MeSH words with radiographs by matching scores obtained by a two-level sparse attention: region sparse attention and word sparse attention. Region sparse attention generates corresponding visual features for each word, and word sparse attention enhances the contribution of images-MeSH matching to the matching scores. To the best of our knowledge, this is the first attempt to learn domain knowledge during pre-training for the medical domain. We evaluate the pre-training model on Radiograph Diagnosis and Reports Generation tasks across four challenging datasets: MIMIC-CXR, IU X-Ray, COV-CTR, and NIH, and achieve state-of-the-art results for all the tasks, which demonstrates the effectiveness of our pre-training model.

poster session 6 @ red 4, poster session 7 @ red 4, oral session 7 @ red 4, poster session 6, poster session 7, oral session 7

The formation of stable coalitions is a central concern in multiagent systems. A considerable stream of research defines stability via the absence of beneficial deviations by single agents. Such deviations require an agent to improve her utility by joining another coalition while possibly imposing further restrictions on the consent of the agents in the welcoming as well as the abandoned coalition. While most of the literature focuses on unanimous consent, we also study consent decided by majority vote, and introduce two new stability notions that can be seen as local variants of popularity. We investigate these notions in additively separable hedonic games by pinpointing boundaries to computational complexity depending on the type of consent and restrictions on the utility functions. The latter restrictions shed new light on well-studied classes of games based on the appreciation of friends or the aversion to enemies. Many of our positive results follow from the Deviation Lemma, a general combinatorial observation, which can be leveraged to prove the convergence of simple and natural single-agent dynamics under fairly general conditions.

poster session 6 @ blue 6, poster session 10 @ blue 6, oral session 10 @ blue 6, poster session 6, poster session 10, oral session 10

The goal of stance detection is to determine the viewpoint expressed in a piece of text towards a target. These viewpoints or contexts are often expressed in many different languages depending on the user and the platform, which can be a local news outlet, a social media platform, a news forum, etc. Most research in stance detection, however, has been limited to working with a single language and on a few limited targets, with little work on cross-lingual stance detection. Moreover, non-English sources of labelled data are often scarce and present additional challenges. Recently, large multilingual language models have substantially improved the performance on many non-English tasks, especially such with limited numbers of examples. This highlights the importance of model pre-training and its ability to learn from few examples. In this paper, we present the most comprehensive study of cross-lingual stance detection to date: we experiment with 15 diverse datasets in 12 languages from 6 language families, and with 6 low-resource evaluation settings each. For our experiments, we build on pattern-exploiting training, proposing the addition of a novel label encoder to simplify the verbalisation procedure. We further propose sentiment-based generation of stance data for pre-training, which shows sizeable improvement of more than 6% F1 absolute in low-shot settings compared to several strong baselines.

poster session 6 @ red 5, poster session 9 @ red 5, oral session 9 @ red 5, poster session 6, poster session 9, oral session 9

Many implicit inferences exist in text depending on how it is structured that can critically impact the text's interpretation and meaning. One such structural aspect present in text with chronology is the order of its presentation. For narratives or stories, this is known as the narrative order. Reordering a narrative can impact the temporal, causal, event-based, and other inferences readers draw from it, which in turn can have strong effects both on its interpretation and interestingness. In this paper, we propose and investigate the task of Narrative Reordering (NAREOR) which involves rewriting a given story in a different narrative order while preserving its plot. We present a dataset, NAREORC, with human rewritings of stories within ROCStories in non-linear orders, and conduct a detailed analysis of it. Further, we propose novel task-specific training methods with suitable evaluation metrics. We perform experiments on NAREORC using state-of-the-art models such as BART and T5 and conduct extensive automatic and human evaluations. We demonstrate that although our models can perform decently, NAREOR is a challenging task with potential for further exploration. We also investigate two applications of NAREOR: generation of more interesting variations of stories and serving as adversarial sets for temporal/event-related tasks, besides discussing other prospective ones, such as for pedagogical setups related to language skills like essay writing and applications to medicine involving clinical narratives.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

Recent neural rendering techniques have greatly benefited image-based modeling and relighting tasks. They provide a continuous, compact, and parallelable representation by modeling the plenoptic function as multilayer perceptrons (MLPs). However, vanilla MLPs suffer from spectral biases on multidimensional datasets. Recent rescues based on isotropic Fourier features mapping mitigate the problem but still fall short of handling heterogeneity across different dimensions, causing imbalanced regression and visual artifacts such as excessive blurs. We present an anisotropic random Fourier features (RFF) mapping scheme to tackle spectral biases. We first analyze the influence of bandwidth from a different perspective: we show that the optimal bandwidth exhibits strong correlations with the frequency spectrum of the training data across various dimensions. We then introduce an anisotropic feature mapping scheme with multiple bandwidths to model the multidimensional signal characteristics. We further propose an efficient bandwidth searching scheme through iterative golden-section search that can significantly reduce the training overload from polynomial time to logarithm. Our anisotropic scheme directly applies to neural surface light-field rendering and image-based relighting. Comprehensive experiments show that our scheme can more faithfully model lighting conditions and object features as well as preserve fine texture details and smooth view transitions even when angular and spatial samples are highly imbalanced.

poster session 6 @ red 3, poster session 10 @ red 3, oral session 10 @ red 3, poster session 6, poster session 10, oral session 10

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

Following up on purely theoretical work of Bredereck et al. [AAAI 2020], we contribute further theoretical insights into adapting stable two-sided matchings to change. Moreover, we perform extensive empirical studies hinting at numerous practically useful properties. Our theoretical extensions include the study of new problems (that is, incremental variants of Almost Stable Marriage and Hospital Residents), focusing on their (parameterized) computational complexity and the equivalence of various change types (thus simplifying algorithmic and complexity-theoretic studies for various natural change scenarios). Our experimental findings reveal, for instance, that allowing the new matching to be blocked by a few pairs significantly decreases the difference between the old and the new matching.

poster session 6 @ blue 6, poster session 10 @ blue 6, poster session 6, poster session 10

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

Though deep learning-based object detection methods have achieved promising results on the conventional datasets, it is still challenging to locate objects from the low-quality images captured in adverse weather conditions. The existing methods either have difficulties in balancing the tasks of image enhancement and object detection, or often ignore the latent information beneficial for detection. To alleviate this problem, we propose a novel Image-Adaptive YOLO (IA-YOLO) framework, where each image can be adaptively enhanced for better detection performance. Specifically, a differentiable image processing (DIP) module is presented to take into account the adverse weather conditions for YOLO detector, whose parameters are predicted by a small convolutional neural network (CNN-PP). We learn CNN-PP and YOLOv3 jointly in an end-to-end fashion, which ensures that CNN-PP can learn an appropriate DIP to enhance the image for detection in a weakly supervised manner. Our proposed IA-YOLO approach can adaptively process images in both normal and adverse weather conditions. The experimental results are very encouraging, demonstrating the effectiveness of our proposed IA-YOLO method in both foggy and low-light scenarios. The source code can be found at https://github.com/wenyyu/Image-Adaptive-YOLO.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

We provide guarantees for approximate Gaussian Process (GP) regression resulting from two common low-rank kernel approximations: based on random Fourier features, and based on truncating the kernel's Mercer expansion. In particular, we bound the Kullback–Leibler divergence between an exact GP and one resulting from one of the afore-described low-rank approximations to its kernel, as well as between their corresponding predictive densities, and we also bound the error between predictive mean vectors and between predictive covariance matrices computed using the exact versus using the approximate GP. We provide experiments on both simulated data and standard benchmarks to evaluate the effectiveness of our theoretical bounds.

poster session 6 @ blue 3, poster session 7 @ blue 3, oral session 7 @ blue 3, poster session 6, poster session 7, oral session 7

Vision transformers (ViTs) have recently received explosive popularity, but the huge computational cost is still a severe issue. Since the computation complexity of ViT is quadratic with respect to the input sequence length, a mainstream paradigm for computation reduction is to reduce the number of tokens. Existing designs include structured spatial compression that uses a progressive shrinking pyramid to reduce the computations of large feature maps, and unstructured token pruning that dynamically drops redundant tokens. However, the limitation of existing token pruning lies in two folds: 1) the incomplete spatial structure caused by pruning is not compatible with structured spatial compression that is commonly used in modern deep-narrow transformers; 2) it usually requires a time-consuming pre-training procedure. To tackle the limitations and expand the applicable scenario of token pruning, we present Evo-ViT, a self-motivated slow-fast token evolution approach for vision transformers. Specifically, we conduct unstructured instance-wise token selection by taking advantage of the simple and effective global class attention that is native to vision transformers. Then, we propose to update the selected informative tokens and uninformative tokens with different computation paths, namely, slow-fast updating. Since slow-fast updating mechanism maintains the spatial structure and information flow, Evo-ViT can accelerate vanilla transformers of both flat and deep-narrow structures from the very beginning of the training process. Experimental results demonstrate that our method significantly reduces the computational cost of vision transformers while maintaining comparable performance on image classification. For example, our method accelerates DeiT-S by over 60% throughput while only sacrificing 0.4% top-1 accuracy on ImageNet-1K, outperforming current token pruning methods on both accuracy and efficiency. Code is available at https://github.com/YifanXu74/Evo-ViT.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

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

It is of significance for an agent to autonomously explore the environment and learn a widely applicable and general-purpose goal-conditioned policy that can achieve diverse goals including images and text descriptions. Considering such perceptually-specific goals, one natural approach is to reward the agent with a prior non-parametric distance over the embedding spaces of states and goals. However, this may be infeasible in some situations, either because it is unclear how to choose suitable measurement, or because embedding (heterogeneous) goals and states is non-trivial. The key insight of this work is that we introduce a latent-conditioned policy to provide goals and intrinsic rewards for learning the goal-conditioned policy. As opposed to directly scoring current states with regards to goals, we obtain rewards by scoring current states with associated latent variables. We theoretically characterize the connection between our unsupervised objective and the multi-goal setting, and empirically demonstrate the effectiveness of our proposed method which substantially outperforms prior techniques in a variety of tasks.

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

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

How do we perform efficient inference while retaining high translation quality? Existing neural machine translation models, such as Transformer, achieve high performance, but they decode words one by one, which is inefficient. Recent non-autoregressive translation models speed up the inference, but their quality is still inferior. In this work, we propose DSLP, a highly efficient and high-performance model for machine translation. The key insight is to train a non-autoregressive Transformer with Deep Supervision and feed additional Layer-wise Predictions. We conducted extensive experiments on four translation tasks (both directions of WMT'14 EN-DE and WMT'16 EN-RO). Results show that our approach consistently improves the BLEU scores compared with respective base models. Specifically, our best variant outperforms the autoregressive model on three translation tasks, while being 14.8 times more efficient in inference.

poster session 6 @ red 5, poster session 9 @ red 5, oral session 6 @ red 5, poster session 6, poster session 9, oral session 6

We propose a new object-centric framework for learning-based stereo 3D object detection. Previous studies build scene-centric representations that do not consider the significant variation among outdoor instances and thus lack the flexibility and functionalities that an instance-level model can offer. We build such an instance-level model by formulating and tackling a local update problem, i.e., how to predict a refined update given an initial 3D cuboid guess. We demonstrate how solving this problem can complement scene-centric approaches in (i) building a coarse-to-fine multi-resolution system, (ii) performing model-agnostic object location refinement, and (iii) conducting stereo 3D tracking-by-detection. Extensive experiments demonstrate the effectiveness of our approach, which achieves state-of-the-art performance on the KITTI benchmark. Code and pre-trained models are available at https://github.com/Nicholasli1995/SNVC.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

Goal-achieving problems are puzzles that set up a specific situation with a clear objective. An example that is well-studied is the category of life-and-death (L&D) problems for Go, which helps players hone their skill of identifying region safety. Many previous methods like lambda search try null moves first, then derive so-called relevance zones (RZs), outside of which the opponent does not need to search. This paper first proposes a novel RZ-based approach, called the RZ-Based Search (RZS), to solving L&D problems for Go. RZS tries moves before determining whether they are null moves post-hoc. This means we do not need to rely on null move heuristics, resulting in a more elegant algorithm, so that it can also be seamlessly incorporated into AlphaZero's super-human level play in our solver. To repurpose AlphaZero for solving, we also propose a new training method called Faster to Life (FTL), which modifies AlphaZero to entice it to win more quickly. We use RZS and FTL to solve L&D problems on Go, namely solving 68 among 106 problems from a professional L&D book while a previous state-of-the-art program TSUMEGO-EXPLORER solves 11 only. Finally, we discuss that the approach is generic in the sense that RZS is applicable to solving many other goal-achieving problems for board games.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

Computerized Adaptive Testing (CAT) refers to an efficient and personalized test mode in online education, aiming to accurately measure student proficiency level on the required subject/domain. The key component of CAT is the "adaptive" question selection algorithm, which automatically selects the best suited question for student based on his/her current estimated proficiency, reducing test length. Existing algorithms rely on some manually designed and pre-fixed informativeness/uncertainty metrics of question for selections, which is labor-intensive and not sufficient for capturing complex relations between students and questions. In this paper, we propose a fully adaptive framework named Neural Computerized Adaptive Testing (NCAT), which formally redefines CAT as a reinforcement learning problem and directly learns selection algorithm from real-world data. Specifically, a bilevel optimization is defined and simplified under CAT's application scenarios to make the algorithm learnable. Furthermore, to address the CAT task effectively, we tackle it as an equivalent reinforcement learning problem and propose an attentive neural policy to model complex non-linear interactions. Extensive experiments on real-world datasets demonstrate the effectiveness and robustness of NCAT compared with several state-of-the-art methods.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

Elkind et al. (AAAI, 2021) introduced a model for deliberative coalition formation, where a community wishes to identify a strongly supported proposal from a space of alternatives, in order to change the status quo. In their model, agents and proposals are points in a metric space, agents' preferences are determined by distances, and agents deliberate by dynamically forming coalitions around proposals that they prefer over the status quo. The deliberation process operates via k-compromise transitions, where agents from k (current) coalitions come together to form a larger coalition in order to support a (perhaps new) proposal, possibly leaving behind some of the dissenting agents from their old coalitions. A deliberation succeeds if it terminates by identifying a proposal with the largest possible support. For deliberation in d dimensions, Elkind et al. consider two variants of their model: in the Euclidean model, proposals and agent locations are points in R^d and the distance is measured according to ||...||_2; and in the hypercube model, proposals and agent locations are vertices of the d-dimensional hypercube and the metric is the Hamming distance. They show that in the continuous model 2-compromises are guaranteed to succeed, but in the discrete model for deliberation to succeed it may be necessary to use k-compromises with k >= d. We complement their analysis by (1) proving that in both models it is hard to find a proposal with a high degree of support, and even a 2-compromise transition may be hard to compute; (2) showing that a sequence of 2-compromise transitions may be exponentially long; (3) strengthening the lower bound on the size of the compromise for the d-hypercube model from d to 2^{\Omega(d)}.

poster session 6 @ blue 6, poster session 10 @ blue 6, oral session 10 @ blue 6, poster session 6, poster session 10, oral session 10

Feature Pyramid Network (FPN) has been widely adopted to exploit multi-scale features for scale variation in object detection. However, intrinsic defects in most of the current methods with FPN make it difficult to adapt to the feature of different geometric objects. To address this issue, we introduce geometric prior into FPN to obtain more discriminative features. In this paper, we propose Geometry-aware Feature Pyramid Network (GaFPN), which consists of Geometry-aware Mapping Module, Geometry-aware Predictor Head.Geometry-aware Mapping Module is proposed to make full use of all pyramid features to obtain better proposal features by the weight-generation subnetwork. The weights generation subnetwork generates fusion weight for each layer proposal features by using the geometric information of the proposal. Geometry-aware Predictor Head introduces geometric prior into predictor head by the embedding generation net-work to strengthen feature representation for classification and regression. Our GaFPN can be easily extended to other two-stage object detectors with feature pyramid and applied to instance segmentation task. The proposed GaFPN significantly improves detection performance compared to baseline detectors with ResNet-50-FPN: +1.9, +2.0, +1.7, +1.3, +0.8 points Average Precision (AP) on Faster-RCNN, Cascade R-CNN, Dynamic R-CNN, SABL, and AugFPN respectively on MS COCO dataset.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

Two major challenges of 3D LiDAR Panoptic Segmentation (PS) are that point clouds of an object are surface-aggregated and thus hard to model the long-range dependency especially for large instances, and that objects are too close to separate each other. Recent literature addresses these problems by time-consuming grouping processes such as dual-clustering, mean-shift offsets and etc., or by bird-eye-view (BEV) dense centroid representation that downplays geometry. However, the long-range geometry relationship has not been sufficiently modeled by local feature learning from the above methods. To this end, we present SCAN, a novel sparse cross-scale attention network to first align multi-scale sparse features with global voxel-encoded attention to capture the long-range relationship of instance context, which is able to boost the regression accuracy of the over-segmented large objects. For the surface-aggregated points, SCAN adopts a novel sparse class-agnostic representation of instance centroids, which can not only maintain the sparsity of aligned features to solve the under-segmentation on small objects, but also reduce the computation amount of the network through sparse convolution. Our method outperforms previous methods by a large margin in the SemanticKITTI dataset for the challenging 3D PS task, achieving 1st place with a real-time inference speed.

poster session 6 @ red 3, poster session 10 @ red 3, oral session 10 @ red 3, poster session 6, poster session 10, oral session 10

Multilingual neural machine translation (MNMT) aims to translate multiple languages with a single model and has been proved successful thanks to effective knowledge transfer among different languages with shared parameters. However, it is still an open question which parameters should be shared and which ones need to be task-specific. Currently, the common practice is to heuristically design or search language-specific modules, which is difficult to find the optimal configuration. In this paper, we propose a novel parameter differentiation based method that allows the model to determine which parameters should be language-specific during training. Inspired by cellular differentiation, each shared parameter in our method can dynamically differentiate into more specialized types. We further define the differentiation criterion as inter-task gradient similarity. Therefore, parameters with conflicting inter-task gradients are more likely to be language-specific. Extensive experiments on multilingual datasets have demonstrated that our method significantly outperforms various strong baselines with different parameter sharing configurations. Further analysis reveals that the parameter sharing configuration obtained by our method correlates well with the linguistic proximities.

poster session 6 @ red 5, poster session 9 @ red 5, oral session 6 @ red 5, poster session 6, poster session 9, oral session 6

We consider applications involving a large set of instances of projecting points to polytopes. We develop an intuition guided by theoretical and empirical analysis to show that when these instances follow certain structures, a large majority of the projections lie on vertices of the polytopes. To do these projections efficiently we derive a vertex-oriented incremental algorithm to project a point onto any arbitrary polytope, as well as give specific algorithms to cater to simplex projection and polytopes where the unit box is cut by planes. Such settings are especially useful in web-scale applications such as optimal matching or allocation problems. Several such problems in internet marketplaces (e-commerce, ride-sharing, food delivery, professional services, advertising, etc.), can be formulated as Linear Programs (LP) with such polytope constraints that require a projection step in the overall optimization process. We show that in the very recent work \cite{basu2020}, the polytopic projection is the most expensive step and our efficient projection algorithms help in gaining massive improvements in performance.

poster session 6 @ blue 4, poster session 10 @ blue 4, poster session 6, poster session 10

Training scene graph classification models requires a large amount of annotated image data. Meanwhile, scene graphs represent relational knowledge that can be modeled from symbolic data such as in texts or knowledge graphs. While image annotation demands extensive labor, collecting textual descriptions of natural scenes requires less effort. In this work, we investigate whether textual scene descriptions can substitute for annotated image data. To this end, we employ a scene graph classification framework that is trained not only from annotated images but also from symbolic data. In our architecture, the symbolic entities are first mapped to their correspondent image-grounded representations and then fed into the relational reasoning pipeline. Even though a structured form of knowledge, such as the form in knowledge graphs is not always available, we can generate it from unstructured texts using a transformer-based language model. We show that by fine-tuning the classification pipeline with the extracted knowledge from texts, we can achieve ~8x more accurate results in scene graph classification, ~3x in object classification, and ~1.5x in predicate classification, compared to the supervised baselines with only 1% of the annotated images.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

Data poisoning and backdoor attacks aim to corrupt a machine learning classifier via modifying, adding, and/or removing some carefully selected training examples, such that the corrupted classifier makes incorrect predictions as the attacker desires. The key idea of state-of-the-art certified defenses against data poisoning and backdoor attacks is to create a \emph{majority vote} mechanism to predict the label of a testing example. Moreover, each voter is a base classifier trained on a subset of the training dataset. Classical simple learning algorithms such as $k$ nearest neighbors (kNN) and radius nearest neighbors (rNN) have intrinsic majority vote mechanisms. In this work, we show that the intrinsic majority vote mechanisms in kNN and rNN already provide certified robustness guarantees against data poisoning and backdoor attacks. Moreover, our evaluation results on MNIST and CIFAR10 show that the intrinsic certified robustness guarantees of kNN and rNN outperform those provided by state-of-the-art certified defenses. Our results serve as standard baselines for future certified defenses against data poisoning and backdoor attacks.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

Psychological concepts can help computational linguists to better model the latent semantic spaces of emotions, and understand the underlying states motivating the sharing or suppressing of emotions. This abstract applies the understanding of agency and social interaction in the happiness semantic space to its role in positive emotion. First, BERT-based fine-tuning yields an expanded seed set to understand the vocabulary of the latent space. Next, results benchmarked against many emotion datasets suggest that the approach is valid, robust, offers an improvement over direct prediction, and is useful for downstream predictive tasks related to psychological states.

poster session 6 @ red 5, poster session 9 @ red 5, poster session 6, poster session 9

Bayesian networks represent relations between variables using a directed acyclic graph (DAG). Learning the DAG is an NP-hard problem and exact learning algorithms are feasible only for small sets of variables. We propose two scalable heuristics for learning DAGs in the linear structural equation case. Our methods learn the DAG by alternating between unconstrained gradient descent-based step to optimize an objective function and solving a maximum acyclic subgraph problem to enforce acyclicity. Thanks to this decoupling, our methods scale up beyond thousands of variables.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

Recent techniques to solve photorealistic style transfer within deep convolutional neural networks (CNNs) generally require intensive training from large-scale datasets, thus having limited applicability and poor generalization ability to unseen images or styles. To overcome this, we propose a novel framework, dubbed Deep Translation Prior (DTP), to accomplish photorealistic style transfer through test-time training on given input image pair with untrained networks, which learns an image pair-specific translation prior and thus yields better performance and generalization. Tailored for such test-time training for style transfer, we present novel network architectures, with two sub-modules of correspondence and generation modules, and loss functions consisting of contrastive content, style, and cycle consistency losses. Our framework does not require offline training phase for style transfer, which has been one of the main challenges in existing methods, but the networks are to be solely learned during test time. Experimental results prove that our framework has a better generalization ability to unseen image pairs and even outperforms the state-of-the-art methods.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

Current Heterogeneous Network Embedding (HNE) models can be roughly divided into two types, i.e., relation-aware and metapath-aware models. However, they either fail to represent the non-pairwise relations in heterogeneous graph, or only capable of capturing local information around target node. In this paper, we propose a metapath based multilevel graph attention networks (MMAN) to jointly learn node embeddings on two substructures, i.e., metapath based graphs and hypergraphs extracted from original heterogeneous graph. Extensive experiments on three benchmark datasets for node classification and node clustering demonstrate the superiority of MMAN over the state-of-the-art works.

poster session 6 @ blue 4, poster session 10 @ blue 4, poster session 6, poster session 10

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

Retrosynthetic planning plays an important role in the field of organic chemistry, which could generate a synthetic route for the target product. The synthetic route is a series of reactions which are started from the available molecules. The most challenging problem in the generation of the synthetic route is the large search space of the candidate reactions. Estimating the cost of candidate reactions has been proved effectively to prune the search space, which could achieve a higher accuracy with the same search iteration. And the estimation of one reaction is comprised of the estimations of all its reactants. So, how to estimate the cost of these reactants will directly influence the quality of results. To get a better performance, we propose a new framework, named GNN-Retro, for retrosynthetic planning problem by combining graph neural networks(GNN) and the latest search algorithm. The structure of GNN in our framework could incorporate the information of neighboring molecules, which will improve the estimation accuracy of our framework. The experiments on the USPTO dataset show that our framework could outperform the state-of-the-art methods with a large margin under the same settings.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

Few-shot learning (FSL) aims to classify images under low-data regimes, where the conventional pooled global feature is likely to lose useful local characteristics. Recent work has achieved promising performances by using deep descriptors. They generally take all deep descriptors from neural networks into consideration while ignoring that some of them are useless in classification due to their limited receptive field, e.g., task-irrelevant descriptors could be misleading and multiple aggregative descriptors from background clutter could even overwhelm the object's presence. In this paper, we argue that a Mutual Nearest Neighbor (MNN) relation should be established to explicitly select the query descriptors that are most relevant to each task and discard less relevant ones from aggregative clutters in FSL. Specifically, we propose Discriminative Mutual Nearest Neighbor Neural Network (DMN4) for FSL. Extensive experiments demonstrate that our method outperforms the existing state-of-the-arts on both fine-grained and generalized datasets.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

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

Activation maximization (AM) refers to the task of generating input examples that maximize the activation of a target class of a classifier, which can be used for class-conditional image generation and model interpretation. A popular class of AM method, GAN-based AM, introduces a GAN pre-trained on a large image set, and performs AM over its input random seed or style embeddings, so that the generated images are natural and adversarial attacks are prevented. Most of these methods would require the image set to contain some images of the target class to be visualized. Otherwise they tend to generate other seen class images that most maximizes the target class activation. In this paper, we aim to tackle the case where information about the target class is completely removed from the image set. This would ensure that the generated images truly reflect the target class information residing in the classifier, not the target class information in the image set, which contributes to a more faithful interpretation technique. To this end, we propose ProbeGAN, a GAN-based AM algorithm capable of generating image classes unseen in the image set. Rather than using a pre-trained GAN, ProbeGAN trains a new GAN with AM explicitly included in its training objective. ProbeGAN consists of a class-conditional generator, a seen-class discriminator, and an all-class unconditional discriminator. It can be shown that such a framework can generate images with the features of the unseen target class, while retaining the naturalness as depicted in the image set. Experiments have shown that ProbeGAN can generate unseen-class images with much higher quality than the baselines. We also explore using ProbeGAN as a model interpretation tool.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

Image-guided depth completion aims to generate dense depth maps with sparse depth measurements and corresponding RGB images. Currently, spatial propagation networks (SPNs) are the most popular affinity-based methods in depth completion, but they still suffer from the representation limitation of the fixed affinity and the over smoothing during iterations. Our solution is to estimate independent affinity matrices in each SPN iteration, but it is over-parameterized and heavy calculation.This paper introduces an efficient model that learns the affinity among neighboring pixels with an attention-based, dynamic approach. Specifically, the Dynamic Spatial Propagation Network (DySPN) we proposed makes use of a non-linear propagation model (NLPM). It decouples the neighborhood into parts regarding to different distances and recursively generates independent attention maps to refine these parts into adaptive affinity matrices. Furthermore, we adopt a diffusion suppression (DS) operation so that the model converges at an early stage to prevent over-smoothing of dense depth. Finally, in order to decrease the computational cost required, we also introduce three variations that reduce the amount of neighbors and attentions needed while still retaining similar accuracy. In practice, our method requires less iteration to match the performance of other SPNs and yields better results overall. DySPN outperforms other state-of-the-art (SoTA) methods on KITTI Depth Completion (DC) evaluation by the time of submission and is able to yield SoTA performance in NYU Depth v2 dataset as well.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

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 similarity of feature representations plays a pivotal role in the success of problems related to domain adaptation. Feature similarity includes both the invariance of marginal distributions and the closeness of conditional distributions given the desired response y (e.g., class labels). Unfortunately, traditional methods always learn such features without fully taking into consideration the information in y, which in turn may lead to a mismatch of the conditional distributions or the mix-up of discriminative structures underlying data distributions. In this work, we introduce the recently proposed von Neumann conditional divergence to improve the transferability across multiple domains. We show that this new divergence is differentiable and eligible to easily quantify the functional dependence between features and y. Given multiple source tasks, we integrate this divergence to capture discriminative information in y and design novel learning objectives assuming those source tasks are observed either simultaneously or sequentially. In both scenarios, we obtain favorable performance against state-of-the-art methods in terms of smaller generalization error on new tasks and less catastrophic forgetting on source tasks (in the sequential setup).

poster session 6 @ red 1, poster session 10 @ red 1, oral session 10 @ red 1, poster session 6, poster session 10, oral session 10

Generating 3D dances from music is an emerged research task that benefits a lot of applications in vision and graphics. Previous works treat this task as sequence generation, however, it is challenging to render a music-aligned long-term sequence with high kinematic complexity and coherent movements. In this paper, we reformulate it by a two-stage process, i.e., a key pose generation and then an in-between parametric motion curve prediction, where the key poses are easier to be synchronized with the music beats and the parametric curves can be efficiently regressed to render fluent rhythm-aligned movements. We named the proposed method as DanceFormer, which includes two cascading kinematics-enhanced transformer-guided networks (called DanTrans) that tackle each stage, respectively. Furthermore, we propose a large-scale music conditioned 3D dance dataset, called PhantomDance, that is accurately labeled by experienced animators rather than reconstruction or motion capture. This dataset also encodes dances as key poses and parametric motion curves apart from pose sequences, thus benefiting the training of our DanceFormer. Extensive experiments demonstrate that the proposed method, even trained by existing datasets, can generate fluent, performative, and music-matched 3D dances that surpass previous works quantitatively and qualitatively. Moreover, the proposed DanceFormer, together with the PhantomDance dataset, are seamlessly compatible with industrial animation software, thus facilitating the adaptation for various downstream applications.

poster session 6 @ red 2, poster session 7 @ red 2, oral session 6 @ red 2, poster session 6, poster session 7, oral session 6

Digital cameras transform sensor RAW readings into RGB images by means of their Image Signal Processor (ISP). Computational photography tasks such as image denoising and colour constancy are commonly performed in the RAW domain, in part due to the inherent hardware design, but also due to the appealing simplicity of noise statistics that result from the direct sensor readings. Despite this, the availability of RAW images is limited in comparison with the abundance and diversity of available RGB data. Recent approaches have attempted to bridge this gap by estimating the RGB to RAW mapping: handcrafted model-based methods that are interpretable and controllable usually require manual parameter fine-tuning, while end-to-end learnable neural networks require large amounts of training data, at times with complex training procedures, and generally lack interpretability and parametric control. Towards addressing these existing limitations, we present a novel hybrid model-based and data-driven ISP that builds on canonical ISP operations and is both learnable and interpretable. Our proposed invertible model, capable of bidirectional mapping between RAW and RGB domains, employs end-to-end learning of rich parameter representations, i.e. dictionaries, that are free from direct parametric supervision and additionally enable simple and plausible data augmentation. We evidence the value of our data generation process by extensive experiments under both RAW image reconstruction and RAW image denoising tasks, obtaining state-of-the-art performance in both. Additionally, we show that our ISP can learn meaningful mappings from few data samples, and that denoising models trained with our dictionary-based data augmentation are competitive despite having only few or zero ground-truth labels.

poster session 6 @ red 3, poster session 10 @ red 3, oral session 10 @ red 3, poster session 6, poster session 10, oral session 10

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

State-of-the-art brain-to-text systems have achieved great success in decoding language directly from brain signals using neural networks. However, current approaches are limited to small closed vocabularies which are far from enough for natural communication. In addition, most of the high-performing approaches require data from invasive devices (e.g., ECoG). In this paper, we extend the problem to open vocabulary Electroencephalography(EEG)-To-Text Sequence-To-Sequence decoding and zero-shot sentence sentiment classification on natural reading tasks. We hypothesis that the human brain functions as a special text encoder and propose a novel framework leveraging pre-trained language models (e.g., BART). Our model achieves a 40.1% BLEU-1 score on EEG-To-Text decoding and a 55.6% F1 score on zero-shot EEG-based ternary sentiment classification, which significantly outperforms supervised baselines. Furthermore, we show that our proposed model can handle data from various subjects and sources, showing great potential for a high-performance open vocabulary brain-to-text system once sufficient data is available.

poster session 6 @ red 5, poster session 9 @ red 5, poster session 6, poster session 9

The question of what kind of convolutional neural network (CNN) structure performs well is fascinating. In this work, we move toward the answer with one more step by connecting zero stability and model performance. Specifically, we found that if the discrete ordinary differential equation is zero stable, its corresponding CNN performs well. We first give the interpretation of zero stability in the context of deep learning and then investigate the performance of existing first- and second-order CNNs under different zero-stable circumstances. Based on the preliminary observation, we provide a higher-order discretization to construct CNNs and then propose a zero-stable network (ZeroSNet). To guarantee zero stability of the ZeroSNet, we first deduce a structure that meets consistency conditions and then give a zero stable region of a training-free parameter. By analyzing the roots of a characteristic equation, we theoretically obtain the optimal coefficients of feature maps. Empirically, we show our results from three aspects: We provide extensive empirical evidence of different widths on different datasets to show that the moduli of the characteristic equation's roots are the keys for the performance of CNNs that require historical features; Our experiments show that ZeroSNet outperforms existing CNNs which is based on high-order discretization; ZeroSNets show better robustness against noises on the input. The source code is available at https://github.com/logichen/ZeroSNet.

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

This paper introduces Ranking Info Noise Contrastive Estimation (RINCE), a new member in the family of InfoNCE losses that preserves a ranked ordering of positive samples. In contrast to the standard InfoNCE loss, which requires a strict binary separation of the training pairs into similar and dissimilar samples, RINCE can exploit information about a similarity ranking for learning a corresponding embedding space. We show that the proposed loss function learns favorable embeddings compared to the standard InfoNCE whenever at least noisy ranking information can be obtained or when the definition of positives and negatives is blurry. We demonstrate this for a supervised classification task with additional superclass labels and noisy similarity scores. Furthermore, we show that RINCE can also be applied to unsupervised training with experiments on unsupervised representation learning from videos. In particular, the embedding yields higher classification accuracy, retrieval rates and performs better on out-of-distribution detection than the standard InfoNCE loss.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

Online learning to rank (OLTR) interactively learns to choose lists of items from a large collection based on certain click models that describe users' click behaviors. Most recent works for this problem focus on the stochastic environment where the item attractiveness is assumed to be invariant during the learning process. In many real-world scenarios, however, the environment could be dynamic or even arbitrarily changing. This work studies the OLTR problem in both stochastic and adversarial environments under the position-based model (PBM). We propose a method based on the follow-the-regularized-leader (FTRL) framework with Tsallis entropy and develop a new self-bounding constraint especially designed for PBM. We prove the proposed algorithm simultaneously achieves $O(\log{T})$ regret in the stochastic environment and $O(m\sqrt{nT})$ regret in the adversarial environment, where $T$ is the number of rounds, $n$ is the number of items and $m$ is the number of positions. We also provide a lower bound of order $\Omega(m\sqrt{nT})$ for adversarial PBM, which matches our upper bound and improves over the state-of-the-art lower bound. The experiments show that our algorithm could simultaneously learn in both stochastic and adversarial environments and is competitive compared to existing methods that are designed for a single environment.

poster session 6 @ red 2, poster session 7 @ red 2, poster session 6, poster session 7

We study the problem of learning a Bayesian network (BN) of a set of variables when structural side information about the system is available. It is well known that learning the structure of a general BN is both computationally and statistically challenging. However, often in many applications, side information about the underlying structure can potentially reduce the learning complexity. In this paper, we develop a recursive constraint-based algorithm that efficiently incorporates such knowledge (i.e., side information) into the learning process. In particular, we study two types of structural side information about the underlying BN: (I) an upper bound on its clique number is known, or (II) it is diamond-free. We provide theoretical guarantees for the learning algorithms, including the worst-case number of tests required in each scenario. As a consequence of our work, we show that bounded treewidth BNs can be learned with polynomial complexity. Furthermore, we evaluate the performance and the scalability of our algorithms in both synthetic and real-world structures and show that they outperform the state-of-the-art structure learning algorithms.

poster session 6 @ red 1, poster session 10 @ red 1, oral session 6 @ red 1, poster session 6, poster session 10, oral session 6

Image-text retrieval plays a central role in bridging vision and language, which aims to reduce the semantic discrepancy between images and texts. Most of existing works rely on refined words and objects representation through the data-oriented method to capture the word-object cooccurrence. Such approaches are prone to ignore the asymmetric action relation between images and texts, that is, the text has explicit action representation (i.e., verb phrase) while the image only contains implicit action information. In this paper, we propose Action-aware Memory-Enhanced embedding (AME) method for image-text retrieval, which aims to emphasize the action information when mapping the images and texts into a shared embedding space. Specifically, we integrate action prediction along with an action-aware memory bank to enrich the image and text features with action-similar text features. The effectiveness of our proposed AME method is verified by comprehensive experimental results on two benchmark datasets.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

Orthogonality regularization has proven effective in improving the precision, convergence speed and the training stability of CNNs. Here, we propose a novel Orthogonal Dictionary Convolution Strategy (ODCS) on CNNs to improve orthogonality effect by optimizing the network architecture and changing the regularized object. Specifically, we remove the nonlinear layer in typical convolution block “Conv(BN) + Nonlinear + Pointwise Conv(BN)”, and only impose orthogonal regularization on the front Conv. The structure, “Conv(BN) + Pointwise Conv(BN)”, is then equivalent to a pair of dictionary and encoding, defined in sparse dictionary learning. Thanks to the exact and efficient representation of signal with dictionaries in low-dimensional projections, our strategy could reduce the superfluous information in dictionary Conv kernels. Meanwhile, the proposed strategy relieves the too strict orthogonality regularization in training, which makes hyper-parameters tuning of model to be more flexible. In addition, our ODCS can modify the state-of-the-art models easily without any extra consumption in inference phase. We evaluate it on a variety of CNNs in small-scale (CIFAR), large-scale (ImageNet) and fine-grained (CUB-200-2011) image classification tasks, respectively. The experimental results show that our method achieve a stable and superior improvement.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

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

Answering complex questions about images is an ambitious goal for machine intelligence, which requires a joint understanding of images, text, and commonsense knowledge, as well as a strong reasoning ability. Recently, multimodal Transformers have made a great progress in the task of Visual Commonsense Reasoning (VCR), by jointly understanding visual objects and text tokens through layers of cross-modality attention. However, these approaches do not utilize the rich structure of the scene and the interactions between objects which are essential in answering complex commonsense questions. We propose a \textbf{S}cene \textbf{G}raph \textbf{E}nhanced \textbf{I}mage-\textbf{T}ext \textbf{L}earning ({\bf SGEITL}) framework to incorporate visual scene graph in commonsense reasoning. In order to exploit the scene graph structure, at the model structure level, we propose a multihop graph transformer for regularizing attention interaction among hops. As for pre-training, a scene-graph-aware pre-training method is proposed to leverage structure knowledge extracted in visual scene graph. Moreover, we introduce a method to train and generate domain relevant visual scene graph using textual annotations in a weakly-supervised manner. Extensive experiments on VCR and other tasks show significant performance boost compared with the state-of-the-art methods, and prove the efficacy of each proposed component.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

In recent years, we have seen significant steps taken in the development of self-driving cars. Multiple companies are starting to roll out impressive systems that work in a variety of settings. These systems can sometimes give the impression that full self-driving is just around the corner and that we would soon build cars without even a steering wheel. However, phasing out of the traditional methods of human-vehicle interaction raises the question of the interaction between the self-driving car and the passenger, and with the increased degree of control and autonomy being given to the AI, makes it that much more critical. Recent works have tried to tackle this issue by allowing the passenger to give commands that refer to specific objects in the visual scene. Nevertheless, this is only half the task as the car should also understand the physical destination of the command, which is what we focus on in this paper. We propose an extension where we annotate the 3D end location that the car needs to reach after executing the given command, and evaluate multiple different baselines on predicting this destination location. From our results, we conclude that our dataset and proposed task are challenging. Additionally, we also propose a model that outperforms the prior works adapted for this particular setting.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

We investigate approval-based committee voting with incomplete information about the approval preferences of voters. We consider several models of incompleteness where each voter partitions the set of candidates into approved, disapproved, and unknown candidates, possibly with ordinal preference constraints among candidates in the latter category. This captures scenarios where voters have not evaluated all candidates and/or it is unknown where voters draw the threshold between approved and disapproved candidates. We study the complexity of some fundamental computational problems for a number of classic approval-based committee voting rules including Proportional Approval Voting and Chamberlin-Courant. These problems include that of determining whether a given set of candidates is a possible or necessary winning committee, whether it forms a committee that possibly or necessarily satisfies representation axioms, and whether a given candidate is possibly or necessarily a member of the winning committee.

poster session 6 @ blue 6, poster session 10 @ blue 6, poster session 6, poster session 10

Exiting simple samples in adaptive multi-exit networks through early modules is an effective way to achieve high computational efficiency. One can observe that deployments of multi-exit architecture on resource-constrained devices are easily limited by high memory footprint of early modules. In this paper, we propose a novel approach named recurrent aggregation operator (ReX), which uses Recurrent Neural Networks (RNNs) to effectively aggregate Intra-patch features within a large receptive field to get delicate local representations, while bypassing large early activations. The resulting model, named ReXNet, can be easily extended to dynamic inference by introducing a novel Consistency-based Early Exit, and the exit criteria are based on the consistency of classification decisions over several modules, rather than the entropy of the prediction distribution. Extensive experiments on two benchmark datasets, i.e., Visual Wake Words, ImageNet-1k, demonstrate that our method consistently reduces the peak RAM and average latency of a wide variety of adaptive models on low-power devices.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

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

In object re-identification (ReID), the development of deep learning techniques often involves model updates and deployment. It is unbearable to re-embedding and re-index with the system suspended when deploying new models. Therefore, backward-compatible representation is proposed to enable ``new'' features to be compared with ``old'' features directly, which means that the database is active when there are both ``new'' and ``old'' features in it. Thus we can scroll-refresh the database or even do nothing on the database to update. The existing backward-compatible methods either require a strong overlap between old and new training data or simply conduct constraints at the instance level. Thus they are difficult in handling complicated cluster structures and are limited in eliminating the impact of outliers in old embeddings, resulting in a risk of damaging the discriminative capability of new features. In this work, we propose a Neighborhood Consensus Contrastive Learning (NCCL) method. With no assumptions about the new training data, we estimate the sub-cluster structures of old embeddings. A new embedding is constrained with multiple old embeddings in both embedding space and discrimination space at the sub-class level. The effect of outliers diminished, as the multiple samples serve as ``mean teachers''. Besides, we also propose a scheme to filter the old embeddings with low credibility, further improving the compatibility robustness. Our method ensures backward compatibility without impairing the accuracy of the new model. And it can even improve the new model's accuracy in most scenarios.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

Knowledge-based visual question answering (VQA) is a vision-language task that requires an artificial agent to correctly answer image-related questions using knowledge that is not presented in the given image. It is not only a more challenging task than regular VQA but also a vital step toward building a general VQA system. Most existing knowledgebased VQA systems process knowledge and image information similarly and ignore the fact that the knowledge base (KB) contains complete information about a triplet, while the extracted image information might be incomplete as the relations between two objects are missing or wrongly detected. In this paper, we propose a novel model named dynamic knowledge memory enhanced multi-step graph reasoning (DMMGR), which performs explicit and implicit reasoning over a key-value knowledge memory module and a spatial-aware image graph, respectively. Specifically, the memory module learns a dynamic knowledge representation and generates a knowledge-aware question representation at each reasoning step. Then, this representation is used to guide a graph attention operator over the spatial-aware image graph. Our model achieves new state-of-the-art accuracy on the KRVQR and FVQA datasets. We also conduct ablation experiments to prove the effectiveness of each component of the the proposed model.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

We present a pose adaptive few-shot learning procedure and a two-stage data interpolation regularization, termed Pose Adaptive Dual Mixup (PADMix), for single-image 3D reconstruction. While augmentations via interpolating feature-label pairs are effective in classification tasks, they fall short in shape predictions potentially due to inconsistencies between interpolated products of two images and volumes when rendering viewpoints are unknown. PADMix targets this issue with two sets of mixup procedures performed sequentially. We first perform an input mixup which, combined with a pose adaptive learning procedure, is helpful in learning 2D feature extraction and pose adaptive latent encoding. The stagewise training allows us to build upon the pose invariant representations to perform a follow-up latent mixup under one-to-one correspondences between features and ground-truth volumes. PADMix significantly outperforms previous literature on few-shot settings over the ShapeNet dataset and sets new benchmarks on the more challenging real-world Pix3D dataset.

poster session 6 @ red 2, poster session 7 @ red 2, poster session 6, poster session 7

Automatic identification of salient aspects from user reviews is especially useful for opinion analysis. There has been significant progress in utilizing weakly supervised approaches, which require only a small set of seed words for training aspect classifiers. However, there is still room for improvement, as these methods do not employ hyperbolic space information, nor do they account for the different latent semantics of individual seed words. We propose HDAE, a hyperbolic disentangled aspect extractor which includes a hyperbolic aspect classifier to infer review segments' aspects, and an aspect-disentangled representation to model the latent semantics of each seed word. Superior performance on two datasets and the embedding visualization demonstrate that \system is a more effective approach to leveraging seed words. We further show the effectiveness of the proposed components by an ablation study and a case study.

poster session 6 @ red 5, poster session 9 @ red 5, poster session 6, poster session 9

Optimistic initialization underpins many theoretically sound exploration schemes in tabular domains; however, in the deep function approximation setting, optimism can quickly disappear if initialized naively. We propose a framework for more effectively incorporating optimistic initialization into reinforcement learning for continuous control. Our approach uses metric information about the state-action space to estimate which transitions are still unexplored, and explicitly maintains the initial Q-value optimism for the corresponding state-action pairs. We also develop methods for efficiently approximating these training objectives, and for incorporating domain knowledge into the optimistic envelope to improve sample efficiency. We empirically evaluate these approaches on a variety of hard exploration problems in continuous control, where our method outperforms existing exploration techniques.

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

While text generated by current vision-language models may be accurate and syntactically correct, it is often general. Recent work has used optical character recognition to supplement visual information with text extracted from an image. In many cases, using text in the image improves the specificity and usefulness of generated text. We contend that vision-language models can benefit from additional information extracted from an image. We modify previous multimodal frameworks to accept relevant information from a number of auxiliary classifiers. In particular, we focus on person names as an additional set of tokens and create a novel image-caption dataset to facilitate captioning with person names. The dataset, Politicians and Athletes in Captions (PAC), consists of captioned images of well-known people in context. By fine-tuning pretrained models with this dataset, we demonstrate a model that can naturally integrate facial recognition tokens into generated text by training on limited data.

poster session 3 @ red 4, poster session 6 @ red 4, poster session 3, poster session 6

Landslides refer to occurrences of massive ground movements due to geological (and meteorological) factors, and can have disastrous impact on property, economy, and even lead to loss of life. The advances of remote sensing provide accurate and continuous terrain monitoring, enabling the study and analysis of land deformation which, in turn, can be used for possible landslides forecast. Prior studies either rely on independent observations for displacement prediction or model static land characteristics without considering the subtle interactions between different locations and the dynamic changes of the surface conditions. We present DyLand -- Dynamic Manifold Learning with Normalizing Flows for Land deformation prediction -- a novel framework for learning dynamic structures of terrain surface and improving the performance of land deformation prediction. DyLand models the spatial connections of InSAR measurements and estimates conditional distributions of deformations on the terrain manifold with a novel normalizing flow-based method. Instead of modeling the stable terrains, it incorporates surface permutations and captures the innate dynamics of the land surface while allowing for tractable likelihood estimates on the manifold. Our extensive evaluations on curated InSAR datasets from continuous monitoring of slopes prone to landslides show that DyLand outperforms existing bechmarking models.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

We propose a versatile framework based on random search, \texttt{Sparse-RS}, for score-based sparse targeted and untargeted attacks in the black-box setting. \texttt{Sparse-RS} does not rely on substitute models and achieves state-of-the-art success rate and query efficiency for multiple sparse attack models: $l_0$-bounded perturbations, adversarial patches, and adversarial frames. The $l_0$-version of untargeted \texttt{Sparse-RS} outperforms all black-box and even all white-box attacks for different models on MNIST, CIFAR-10, and ImageNet. Moreover, our untargeted \texttt{Sparse-RS} achieves very high success rates even for the challenging settings of $20\times20$ adversarial patches and $2$-pixel wide adversarial frames for $224\times224$ images. Finally, we show that \texttt{Sparse-RS} can be applied to generate targeted universal adversarial patches where it significantly outperforms the existing approaches.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

Many planning, scheduling or multi-dimensional packing problems involve the design of subtle logical combinations of temporal or spatial constraints. On the one hand, the precise modelling of these constraints, which are formulated in various relation algebras, entails a number of possible logical combinations and requires expertise in constraint-based modelling. On the other hand, active constraint acquisition (CA) has been used successfully to support non-experienced users in learning conjunctive constraint networks through the generation of a sequence of queries. In this paper, we propose GEACQ, which stands for Generic Qualitative Constraint Acquisition, an active CA method that learns qualitative constraints via the concept of qualitative queries. GEACQ combines qualitative queries with time-bounded path consistency (PC) and background knowledge propagation to acquire the qualitative constraints of any scheduling or packing problem. We prove soundness, completeness and termination of GEACQ by exploiting the jointly exhaustive and pairwise disjoint property of qualitative calculus and we give an experimental evaluation that shows (i) the efficiency of our approach in learning temporal constraints and, (ii) the use of GEACQ on real scheduling instances.

poster session 6 @ blue 4, poster session 10 @ blue 4, oral session 10 @ blue 4, poster session 6, poster session 10, oral session 10

Influence diagrams have recently been used to analyse the safety and fairness properties of AI systems. A key building block for this analysis is a graphical criterion for value of information (VoI). This paper establishes the first complete graphical criterion for VoI in influence diagrams with multiple decisions. Along the way, we establish two important techniques for proving properties of multi-decision influence diagrams: ID homomorphisms are structure-preserving transformations of influence diagrams, while a Tree of Systems is a collection of paths that captures how information and control can flow in an influence diagram.

poster session 6 @ red 1, poster session 10 @ red 1, poster session 6, poster session 10

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

Infographics deal with representing data or information visually in a perceptually compelling manner. Recently, infographics have gained widespread popularity, giving rise to automated infographics synthesis from texts. Our research follows an ontological approach to automatically extract the necessary indicators from an input sentence and synthesize an infographic corresponding to it. This work includes (1) the creation of a dataset, (2) an end-to-end domain-agnostic framework, and (3) demonstrating the application of the proposed framework. The results demonstrate our framework's ability to extract the necessary textual cues from real-world textual descriptions (from various domains) and synthesize meaningful infographics.

poster session 6 @ red 5, poster session 9 @ red 5, poster session 6, poster session 9

Text-based image captioning (TextCap) requires simultaneous comprehension of visual content and reading the text of images to generate a natural language description. Although a task can teach machines to understand the complex human environment further given that text is omnipresent in our daily surroundings, it poses additional challenges in normal captioning. A text-based image intuitively contains abundant and complex multimodal relational content, that is, image details can be described diversely from multiview rather than a single caption. Certainly, we can introduce additional paired training data to show the diversity of images' descriptions, this process is labor-intensive and time-consuming for TextCap pair annotations with extra texts. Based on the insight mentioned above, we investigate how to generate diverse captions that focus on different image parts using an unpaired training paradigm. We propose the Multimodal relAtional Graph adversarIal InferenCe (MAGIC) framework for diverse and unpaired TextCap. This framework can adaptively construct multiple multimodal relational graphs of images and model complex relationships among graphs to represent descriptive diversity. Moreover, a cascaded generative adversarial network is developed from modeled graphs to infer the unpaired caption generation in image–sentence feature alignment and linguistic coherence levels. We validate the effectiveness of MAGIC in generating diverse captions from different relational information items of an image. Experimental results show that MAGIC can generate very promising outcomes without using any image–caption training pairs.

poster session 6 @ red 4, poster session 7 @ red 4, oral session 6 @ red 4, poster session 6, poster session 7, oral session 6

Singing voice synthesis (SVS) systems are built to synthesize high-quality and expressive singing voice, in which the acoustic model generates the acoustic features (\eg, mel-spectrogram) given a music score. Previous singing acoustic models adopt a simple loss (\eg, L1 and L2) or generative adversarial network (GAN) to reconstruct the acoustic features, while they suffer from over-smoothing and unstable training issues respectively, which hinder the naturalness of synthesized singing. In this work, we propose DiffSinger, an acoustic model for SVS based on the diffusion probabilistic model. DiffSinger is a parameterized Markov chain that iteratively converts the noise into mel-spectrogram conditioned on the music score. By implicitly optimizing variational bound, DiffSinger can be stably trained and generate realistic outputs. To further improve the voice quality and speed up inference, we introduce a shallow diffusion mechanism to make better use of the prior knowledge learned by the simple loss. Specifically, DiffSinger starts generation at a shallow step smaller than the total number of diffusion steps, according to the intersection of the diffusion trajectories of the ground-truth mel-spectrogram and the one predicted by a simple mel-spectrogram decoder. Besides, we propose boundary prediction methods to locate the intersection and determine the shallow step adaptively. The evaluations conducted on a Chinese singing dataset demonstrate that DiffSinger outperforms state-of-the-art SVS work. Extensional experiments also prove the generalization of our methods on text-to-speech task (DiffSpeech). Audio samples are available via \url{https://diffsinger.github.io}.

poster session 6 @ red 5, poster session 9 @ red 5, poster session 6, poster session 9

We study the performance of voting mechanisms from a utilitarian standpoint, under the recently introduced framework of metric-distortion, offering new insights along two main lines. First, if $d$ represents the doubling dimension of the metric space, we show that the distortion of STV is $O(d \log \log m)$, where $m$ represents the number of candidates. For doubling metrics this implies an exponential improvement over the lower bound for general metrics, and as a special case it effectively answers a question left open by Skowron and Elkind (AAAI '17) regarding the distortion of STV under low-dimensional Euclidean spaces. More broadly, this constitutes the first nexus between the performance of any voting rule and the ``intrinsic dimensionality'' of the underlying metric space. We also establish a nearly-matching lower bound, refining the construction of Skowron and Elkind. Moreover, motivated by the efficiency of STV, we investigate whether natural learning rules can lead to low-distortion outcomes. Specifically, we introduce simple, deterministic and decentralized exploration/exploitation dynamics, and we show that they converge to a candidate with $O(1)$ distortion.

poster session 6 @ blue 6, poster session 10 @ blue 6, poster session 6, poster session 10

Answer Set Programming (ASP) is a well-known declarative AI formalism for knowledge representation and reasoning. State-of-the-art ASP implementations employ the ground&solve approach, and they were successfully applied to industrial and academic problems. Nonetheless there are classes of ASP programs whose evaluation is not efficient (sometimes not feasible) due to the combinatorial blow-up of the program produced by the grounding step. Recent researches suggest that compilation-based techniques can mitigate the grounding bottleneck problem. However, no compilation-based technique has been developed for ASP programs that contain aggregates, which are one of the most relevant and commonly-employed constructs of ASP. In this paper, we propose a compilation-based approach for ASP programs with aggregates. We implement it on top of a state-of-the-art ASP system, and evaluate the performance on publicly-available benchmarks. Experiments show our approach is effective on ground-intensive ASP programs.

poster session 6 @ red 2, poster session 7 @ red 2, oral session 7 @ red 2, poster session 6, poster session 7, oral session 7

This work aims to address the long-established problem of learning diversified representations. To this end, we combine information-theoretic arguments with stochastic competition-based activations, namely Stochastic Local Winner-Takes-All (LWTA) units. In this context, we ditch the conventional deep architectures commonly used in Representation Learning, that rely on non-linear activations; instead, we replace them with sets of locally and stochastically competing linear units. In this setting, each network layer yields sparse outputs, determined by the outcome of the competition between units that are organized into blocks of competitors. We adopt stochastic arguments for the competition mechanism, which perform posterior sampling to determine the winner of each block. We further endow the considered networks with the ability to infer the sub-part of the network that is essential for modeling the data at hand; we impose appropriate stick-breaking priors to this end. To further enrich the information of the emerging representations, we resort to information-theoretic principles, namely the Information Competing Process (ICP). Then, all the components are tied together under the stochastic Variational Bayes framework for inference. We perform a thorough experimental investigation for our approach using benchmark datasets on image classification. As we experimentally show, the resulting networks yield significant discriminative representation learning abilities. In addition, the introduced paradigm allows for a principled investigation mechanism of the emerging intermediate network representations.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

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

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

Multi-class classifier training using traditional meta-algorithms such as the popular One-vs-One (OvO) method may not always work well under cost-sensitive setups. Also, during inference, OvO becomes computationally challenging for higher class counts K as O(K^2) is its time complexity. In this paper, we present Opt-OvO, an optimized (resource-friendly) version of the One-vs-One algorithm to enable high-performance multi-class ML classifier training and inference directly on microcontroller units (MCUs). Opt-OvO enables billions of tiny IoT devices to self learn/train (offline) after their deployment, using live data from a wide range of IoT use-cases. We demonstrate Opt-OvO by performing live ML model training on 4 popular MCU boards using datasets of varying class counts, sizes, and feature dimensions. The most exciting finding was, on the 3 $ ESP32 chip, Opt-OvO trained a multi-class ML classifier using a dataset of class count 50 and performed unit inference in super real-time of 6.2 ms.

poster session 6 @ blue 4, poster session 10 @ blue 4, poster session 6, poster session 10

Gradient temporal difference (Gradient TD) algorithms are a popular class of stochastic approximation (SA) algorithms used for policy evaluation in reinforcement learning. Here, we consider Gradient TD algorithms with an additional heavy ball momentum term and provide choice of step size and momentum parameter that ensures almost sure convergence of these algorithms asymptotically. In doing so, we decompose the heavy ball Gradient TD iterates into three separate iterates with different step sizes. We first analyze these iterates under one-timescale SA setting using results from current literature. However, the one-timescale case is restrictive and a more general analysis can be provided by looking at a three-timescale decomposition of the iterates. In the process we provide the first conditions for stability and convergence of general three-timescale SA. We then prove that the heavy ball Gradient TD algorithm is convergent using our three-timescale SA analysis. Finally, we evaluate these algorithms on standard RL problems and report improvement in performance over the vanilla algorithms.

poster session 6 @ red 2, poster session 7 @ red 2, poster session 6, poster session 7

Recently, automated surveillance cameras can change a visible sensor and a thermal sensor for all-day operation. However, existing single-modal pedestrian detectors mainly focus on detecting pedestrians in only one specific modality (i.e., visible or thermal), so they cannot effectively cope with other modal inputs. In addition, recent multispectral pedestrian detectors have shown remarkable performance by adopting multispectral modalities, but they also have limitations in practical applications (e.g., different Field-of-View (FoV) and frame rate). In this paper, we introduce a versatile pedestrian detector that shows robust detection performance in any single modality. We propose a multisensory-matching contrastive loss to reduce the difference between the visual representation of pedestrians in the visible and the thermal modalities. Moreover, to make the proposed method perform robust detection on a single modality, we design a Multispectral Recalling (MSR) Memory. The MSR Memory enhances the visual representation of the single modal features by recalling that of the multispectral modalities. To guide the MSR Memory to store the contexts of the multispectral modalities, we introduce a multispectral recalling loss. It enables the pedestrian detector to encode more discriminative features with a single input modality. We would like to insist that our method is a step forward detector that can be applied to a variety of real-world applications. The comprehensive experimental results verify the effectiveness of the proposed method.

poster session 6 @ red 3, poster session 10 @ red 3, oral session 6 @ red 3, poster session 6, poster session 10, oral session 6

Financial assets exhibit dependence structures, i.e., movements of their prices or returns show various correlations. Knowledge of assets’ price dependencies can help investors to create a diversified portfolio, aiming to reduce portfolio risk due to the high volatility of the financial market. Since asset dependency changes with time in complex patterns, asset dependency forecast is an essential problem in finance. In this paper, we organize pairwise assets dependencies in an Asset Dependency Matrix (ADM) and formulate the problem of assets dependencies forecast to predict the future ADM given a sequence of past ADMs. We propose a novel idea viewing a sequence of ADMs as a sequence of images to capture the spatial and temporal dependencies among the assets. Inspired by video prediction tasks, we develop a novel Asset Dependency Neural Network (ADNN) to tackle the ADM prediction problem. Experiments show that our proposed framework consistently outperforms baselines on both future ADM prediction and portfolio risk reduction tasks.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

Since people know the important factors that affect the control based on the driving situation, they can drive safely even in diverse driving environment. To mimic these behaviors, we propose a two-stage representation learning model that initially splits the latent features as domain-specific features, which contain information in a specific domain, and domain-general features, which are consistent across all domains. Subsequently, the dynamic-object features, which contain information of dynamic objects, are disentangled from latent features using mutual information estimator. In this study, the problem in behavior cloning is divided into several domain-specific subspaces, with experts becoming specialized on each domain-specific policy. The proposed mixture of domain-specific experts (MoDE) model predicts the final control values through the cooperation of experts using a gating function. The domain-specific features are used to calculate the importance weight of the domain-specific experts, and the disentangled domain-general and dynamic-object features are applied in estimating the control values. To validate the proposed MoDE model, we conducted several experiments and achieved a higher success rate on the CARLA benchmarks under several conditions and tasks than state-of-the-art approaches.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

Fine-grained visual classification (FGVC) which aims at recognizing objects from subcategories is a very challenging task due to the inherently subtle inter-class differences. Most existing works mainly tackle this problem by reusing the backbone network to extract features of detected discriminative regions. However, this strategy inevitably complicates the pipeline and pushes the proposed regions to contain most parts of the objects thus fails to locate the really important parts. Recently, vision transformer (ViT) shows its strong performance in the traditional classification task. The self-attention mechanism of the transformer links every patch token to the classification token. In this work, we first evaluate the effectiveness of the ViT framework in the fine-grained recognition setting. Then motivated by the strength of the attention link can be intuitively considered as an indicator of the importance of tokens, we further propose a novel Part Selection Module that can be applied to most of the transformer architectures where we integrate all raw attention weights of the transformer into an attention map for guiding the network to effectively and accurately select discriminative image patches and compute their relations. A contrastive loss is applied to enlarge the distance between feature representations of confusing classes. We name the augmented transformer-based model TransFG and demonstrate the value of it by conducting experiments on five popular fine-grained benchmarks where we achieve state-of-the-art performance. Qualitative results are presented for better understanding of our model.

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

Uplift modeling aims to model the incremental impact of a treatment on an individual outcome, which has attracted great interests of researchers and practitioners from different communities.Existing uplift modeling methods rely on either the data collected from randomized controlled trials (RCTs) or the observational data which is more realistic. However, we notice that on the observational data, it is often the case that only a small number of subjects receive treatment, but finally infer the uplift on a much large group of subjects. Such highly imbalanced data is common in various fields such as marketing and medical treatment but it is rarely handled by existing works. In this paper, we theoretically and quantitatively prove that the existing representative methods, transformed outcome (TOM) and doubly robust (DR), suffer from large bias and deviation on highly imbalanced datasets with skewed propensity scores, mainly because they are proportional to the reciprocal of the propensity score. To reduce the bias and deviation of uplift modeling with an imbalanced dataset, we propose an imbalance-aware uplift modeling (IAUM) method via constructing a robust proxy outcome, which adaptively combines the doubly robust estimator and the imputed treatment effects based on the propensity score. We theoretically prove that IAUM can obtain a better bias-variance trade-off than existing methods on a highly imbalanced dataset. We conduct extensive experiments a synthetic dataset and a real-world dataset collected from a marketing campaign. The experimental results well demonstrate the superiority of our method over state-of-the-art.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

Unconstrained lip-to-speech aims to generate corresponding speeches based on silent facial videos with no restriction to head pose or vocabulary. It is desirable to generate intelligible and natural speech with a fast speed in unconstrained settings. Currently, to handle the more complicated scenarios, most existing methods adopt the autoregressive architecture, which are optimized with the MSE loss. Although these methods have achieved promising performance, they are prone to bring issues including high inference latency and mel-spectrogram over-smoothness. To tackle these problems, we propose a novel flow-based non-autoregressive lip-to-speech model (GlowLTS) to break autoregressive constraints and achieve faster inference. Concretely, we adopt a flow-based decoder which is optimized by maximizing the likelihood of the training data and is capable of more natural and fast speech generation. Moreover, we devise a condition module to improve the intelligibility of generated speech. We demonstrate the superiority of our proposed method through objective and subjective evaluation on Lip2Wav-Chemisty-Lectures and Lip2Wav-Chess-Analysis datasets. Our demo video can be found at \url{https://glowlts.github.io/}.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

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

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

Federated learning (FL) performs the global model updating in a synchronous manner in that the FL server waits for a specific number of local models from distributed devices before computing and sharing a new global model. We propose asynchronous federated learning (AsyncFL), which allows each client to continuously upload its model based on its capabilities and the FL server to determine when to asynchronously update and broadcast the global model. The asynchronous model aggregation at the FL server is performed by the Boyer–Moore majority voting algorithm for the k-bit quantized weight values. The proposed FL can speed up the convergence of the global model learning early in the FL process and reduce data exchange once the model is converged.

poster session 6 @ blue 6, poster session 10 @ blue 6, poster session 6, poster session 10

Entropy estimation is an important problem in information theory and statistical science. Many popular entropy estimators suffer from fast growing estimation bias with respect to dimensionality, rendering them unsuitable for high dimensional problems. In this work we propose a transformbased method for high dimensional entropy estimation, which consists of the following two main ingredients. First by modifying the k-NN based entropy estimator developed in (Kozachenko & Leonenko, 1987), we propose a new estimator which enjoys small estimation bias for samples that are close to a uniform distribution. Second we design a normalizing flow based mapping that pushes samples toward a uniform distribution, and the relation between the entropy of the original samples and the transformed ones is also derived. As a result the entropy of a given set of samples is estimated by first transforming them toward a uniform distribution and then applying the proposed estimator to the transformed samples. Numerical experiments demonstrate the effectiveness of the method for high dimensional entropy estimation problems.

poster session 6 @ red 1, poster session 10 @ red 1, oral session 6 @ red 1, poster session 6, poster session 10, oral session 6

Continual learning is an intellectual ability of artificial agents to learn new concepts from sequential data. The main impediment to continual learning is catastrophic forgetting, a severe performance degradation on previously learned tasks. Although simply replaying all previous data or continuously adding the model parameters could alleviate the issue, it is impractical in real-world applications due to the limited available resources. Inspired by the mechanism of the human brain to deepen its past impression, we propose a novel framework, Deep Retrieval and Imagination (DRI), which consists of two components: 1) an embedding network that constructs a unified embedding space without adding model parameters on the arrival of new tasks; and 2) a generative model to produce additional (imaginary) data based on the limited memory. By retrieving the past experiences and corresponding imaginary data, DRI distills knowledge and rebalances the embedding space to further mitigate forgetting. Theoretical analysis demonstrates that DRI can reduce the loss approximation error and improve the robustness through retrieval and imagination, bringing better generalizability to the network. Extensive experiments show that DRI performs significantly better than the existing state-of-the-art continual learning methods and effectively alleviates catastrophic forgetting.

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

We present a customized 3D mesh Transformer model for the pose transfer task. As the 3D pose transfer essentially is a deformation procedure dependent on the given meshes, the intuition of this work is to perceive the geometric inconsistency between the given meshes with the powerful self-attention mechanism. Specifically, we propose a novel geometry-contrastive Transformer that has an efficient 3D structured perceiving ability to the global geometric inconsistencies across the given meshes. Moreover, locally, a simple yet efficient central geodesic contrastive loss is further proposed to improve the regional geometric-inconsistency learning. At last, we present a latent isometric regularization module together with a novel semi-synthesized dataset for the cross-dataset 3D pose transfer task towards unknown spaces. The massive experimental results prove the efficacy of our approach by showing state-of-the-art quantitative performances on SMPL-NPT, FAUST and our new proposed dataset SMG-3D datasets, as well as promising qualitative results on MG-cloth and SMAL datasets. It's demonstrated that our method can achieve robust 3D pose transfer and be generalized to challenging meshes from unknown spaces on cross-dataset tasks. The code and dataset are made available. Code is available: https://github.com/mikecheninoulu/CGT.

poster session 6 @ red 2, poster session 7 @ red 2, poster session 6, poster session 7

Vehicle re-identification (ReID) has attracted considerable attention in computer vision. Although several methods have been proposed to achieve state-of-the-art performance on this topic, re-identifying vehicle in foggy scenes remains a great challenge due to the degradation of visibility. To our knowledge, this problem is still not well-addressed so far. In this paper, to address this problem, we proposed a novel training framework called Semi-supervised Joint Defogging Learning (SJDL) framework. First, the fog removal branch and the re-identification branch are integrated to perform simultaneous training. With the collaborative training scheme, defogged features generated by the defogging branch from input images can be shared to learn better representation for the re-identification branch. However, since the fog-free image of real-world data is intractable, this architecture can only be trained on the synthetic data, which may cause the domain gap problem between real-world and synthetic scenarios. To solve this problem, we design a semi-supervised defogging training scheme which can train two kinds of data alternatively in each iteration. Due to the lack of dataset specialized for vehicle ReID in foggy weather, we construct a dataset called FVRID which consists of real-world and synthetic foggy images to train and evaluate the performance. Experimental results show that the proposed method is effective and outperforms other existing vehicle ReID methods in foggy weather.

poster session 6 @ red 2, poster session 7 @ red 2, poster session 6, poster session 7

The practical scalability of many optimization algorithms for large extensive-form games is often limited by the games' huge payoff matrices. To ameliorate the issue, Zhang and Sandholm recently proposed a sparsification technique that factorizes the payoff matrix $\mathbf{A}$ into a sparser object $\mathbf{A} = \hat{\mathbf{A}} + \mathbf{U}\mathbf{V}^\top$, where the total combined number of nonzeros of $\hat{\mathbf{A}}$, $\mathbf{U}$, and $\mathbf{V}$, is significantly smaller. Such a factorization can be used in place of the original payoff matrix in many optimization algorithm, such as interior-point and second-order methods, thus increasing the size of games that can be handled. Their technique significantly sparsifies poker (end)games, standard benchmarks used in computational game theory, AI, and more broadly. We show that the existence of extremely sparse factorizations in poker games can be tied to their particular Kronecker-product structure. We clarify how such structure arises and introduce the connection between that structure and sparsification. By leveraging such structure, we give two ways of computing strong sparsifications of poker games (as well as any other game with a similar structure) that are i) orders of magnitude faster to compute, ii) more numerically stable, and iii) produce a dramatically smaller number of nonzeros than the prior technique. Our techniques enable---for the first time---effective computation of high-precision Nash equilibria and strategies subject to constraints on the amount of allowed randomization. Furthermore, they significantly speed up parallel first-order game-solving algorithms; we show state-of-the-art speed on a GPU.

poster session 6 @ blue 6, poster session 10 @ blue 6, poster session 6, poster session 10

Finding diverse solutions in combinatorial problems recently has received considerable attention (Baste et al. 2020, Fomin et al. 2020, Hanaka et al. 2021). In this paper we study the following type of problems: given an integer $k$, the problem asks for $k$ solutions such that the sum of pairwise (weighted) Hamming distances between these solutions is maximized. Such solutions are called \emph{diverse solutions}. We present a polynomial-time algorithm for finding diverse shortest $st$-paths in weighted directed graphs. Moreover, we study the diverse version of other classical combinatorial problems such as diverse weighted matroid bases, diverse weighted arborescences, and diverse bipartite matchings. We show that these problems can be solved in polynomial time as well. To evaluate the practical performance of our algorithm for finding diverse shortest $st$-paths, we conduct a computational experiment with synthetic and real-world instances. The experiment shows that our algorithm successfully computes diverse solutions within reasonable computational time.

poster session 6 @ blue 4, poster session 10 @ blue 4, oral session 6 @ blue 4, poster session 6, poster session 10, oral session 6

As an asset pricing model in economics and finance, factor model has been widely used in quantitative investment. Towards building more effective factor models, recent years have witnessed the paradigm shift from linear models to more flexible nonlinear data-driven machine learning models. However, due to low signal-to-noise ratio of the financial data, it is quite challenging to learn effective factor models. In this paper, we propose a novel factor model, FactorVAE, as a probabilistic model with inherent randomness for noise modeling. Essentially, our model integrates the dynamic factor model (DFM) with the variational autoencoder (VAE) in machine learning, and we propose a prior-posterior learning method based on VAE, which can effectively guide the learning of model by approximating an optimal posterior factor model with future information. Particularly, considering that risk modeling is important for the noisy stock data, FactorVAE can estimate the variances from the distribution over the latent space of VAE, in addition to predicting returns. The experiments on the real stock market data demonstrate the effectiveness of FactorVAE, which outperforms various baseline methods.

poster session 6 @ red 6, poster session 11 @ red 6, poster session 6, poster session 11

Belief revision aims at incorporating, in a rational way, a new piece of information into the beliefs of an agent. Most works in belief revision suppose a classical logic setting, where the beliefs of the agent are consistent. Moreover, the consistency postulate states that the result of the revision should be consistent if the new piece of information is consistent. But in real applications it may easily happen that (some parts of) the beliefs of the agent are not consistent. In this case then it seems reasonable to use paraconsistent logics to derive sensible conclusions from these inconsistent beliefs. However, in this context, the standard belief revision postulates trivialize the revision process. In this work we discuss how to adapt these postulates when the underlying logic is Priest's LP logic, in order to model a rational change, while being a conservative extension of AGM/KM belief revision. This implies, in particular, to adequately adapt the notion of expansion. We provide a representation theorem and some examples of belief revision operators in this setting.

poster session 6 @ red 2, poster session 7 @ red 2, poster session 6, poster session 7

Using natural-language feedback to guide image generation and manipulation can greatly lower the required efforts and skills. This topic has received increased attention in recent years through refinement of Generative Adversarial Networks (GANs); however, most existing works are limited to single-round interaction, which is not reflective of real world interactive image editing workflows. Furthermore, previous works dealing with multi-round scenarios are limited to predefined feedback sequences, which is also impractical. In this paper, we propose a novel framework for Text-based Interactive image generation and manipulation (TiGAN) that responds to users' natural-language feedback. TiGAN utilizes the powerful pre-trained CLIP model to understand users' natural-language feedback and exploits contrastive learning for a better text-to-image mapping. To maintain the image consistency during interactions, TiGAN generates intermediate feature vectors aligned with the feedback and selectively feeds these vectors to our proposed generative model. Empirical results on several datasets show that TiGAN improves both interaction efficiency and image quality while better avoids undesirable image manipulation during interactions.

poster session 6 @ red 4, poster session 7 @ red 4, poster session 6, poster session 7

We study here a way to approximate information retrieval metrics through a softmax-based approximation of the rank indicator function. Indeed, this latter function is a key component in the design of information retrieval metrics, as well as in the design of the ranking and sorting functions. Obtaining a good approximation for it thus opens the door to differentiable approximations of many evaluation measures that can in turn be used in neural end-to-end approaches. We first prove theoretically that the approximations proposed are of good quality, prior to validate them experimentally on both learning to rank and text-based information retrieval tasks.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

We attempt to train deep neural networks for classification without using any labeled data. Existing unsupervised methods, though mine useful clusters or features, require some annotated samples to facilitate the final task-specific predictions. This defeats the true purpose of unsupervised learning and hence we envisage a paradigm of `true' self-supervision, where absolutely no annotated instances are used for training a classifier. The proposed method first pretrains a deep network through self-supervision and performs clustering on the learned features. A classifier layer is then appended to the self-supervised network and is trained by matching the distribution of the predictions to that of a predefined prior. This approach leverages the distribution of labels for supervisory signals and consequently, no image-label pair is needed. Experiments reveal that the method works on major nominal as well as ordinal classification datasets and delivers significant performance.

poster session 6 @ red 3, poster session 10 @ red 3, poster session 6, poster session 10

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

Recently, contrastive learning has largely advanced the progress of unsupervised visual representation learning. Pre-trained on ImageNet, some self-supervised algorithms reported higher transfer learning performance compared to fully-supervised methods, seeming to deliver the message that human labels hardly contribute to learning transferrable visual features. In this paper, we defend the usefulness of semantic labels but point out that fully-supervised and self-supervised methods are pursuing different kinds of features. To alleviate this issue, we present a new algorithm named Supervised Contrastive Adjustment in Neighborhood (SCAN) that maximally prevents the semantic guidance from damaging the appearance feature embedding. In a series of downstream tasks, SCAN achieves superior performance compared to previous fully-supervised and self-supervised methods, and sometimes the gain is significant. More importantly, our study reveals that semantic labels are useful in assisting self-supervised methods, opening a new direction for the community.

poster session 6 @ red 3, poster session 10 @ red 3, oral session 6 @ red 3, poster session 6, poster session 10, oral session 6

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

Over the last few years, researchers have put significant effort into understanding of the notion of proportional representation in committee election. In particular, recently they have proposed the notion of proportionality degree. We study the complexity of computing committees with a given proportionality degree and of testing if a given committee provides a particular one. This way, we complement recent studies that mostly focused on the notion of (extended) justified representation. We also study the problems of testing if a cohesive group of a given size exists and of counting such groups.

poster session 6 @ blue 6, poster session 10 @ blue 6, poster session 6, poster session 10

How to get a good value estimation is one of the key problems in reinforcement learning (RL). Current off-policy methods, such as Maxmin Q-learning, TD3 and TADD, suffer from the underestimation problem when solving the overestimation problem. In this paper, we propose the Quasi-Median Operation, a novel way to mitigate the underestimation bias by selecting the quasi-median from multiple state-action values. Based on the quasi-median operation, we propose Quasi-Median Q-learning (QMQ) for the discrete action tasks and Quasi-Median Delayed Deep Deterministic Policy Gradient (QMD3) for the continuous action tasks. Theoretically, the underestimation bias of our method is improved while the estimation variance is significantly reduced compared to Maxmin Q-learning, TD3 and TADD. We conduct extensive experiments on the discrete and continuous action tasks, and results show that our method outperforms the state-of-the-art methods.

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

Meta-learning models transfer the knowledge acquired from previous tasks to quickly learn new ones. They are trained on benchmarks with a fixed number of data points per task. This number is usually arbitrary and it is unknown how it affects performance at testing. Since the labelling of data is expensive, finding the optimal allocation of labels across training tasks may reduce costs. Given a fixed budget of labels, should we use a small number of highly labelled tasks or many tasks with few labels each? Should we allocate more labels to some tasks and less to others? We show that: 1) If tasks are homogeneous, there is a uniform optimal allocation, whereby all tasks get the same amount of data; 2) At fixed budget, there is a trade-off between number of tasks and number of data points per task, with a unique and constant optimum; 3) When trained separately, the harder task should get more data, at the cost of a smaller number of tasks; 4) When training on a mixture of easy and hard tasks, more data should be allocated to easy tasks. Interestingly, Neuroscience experiments have shown that human visual skills also transfer better from easy tasks. We prove these results mathematically on mixed linear regression, and we show empirically that the same results hold for few-shot image classification on CIFAR-FS and mini-ImageNet. Our results provide guidance for allocating labels across tasks when collecting data for meta-learning.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

While Answer-Set Programming (ASP) is a prominent approach to declarative problem solving, optimisation problems can still be a challenge for it. Large-Neighbourhood Search (LNS) is a metaheuristic for optimisation where parts of a solution are alternately destroyed and reconstructed that has high but untapped potential for ASP solving. We present a framework for LNS optimisation in answer-set solving, in which neighbourhoods can be specified either declaratively as part of the ASP encoding, or automatically generated by code. To effectively explore different neighbourhoods, we focus on multi-shot solving as it allows to avoid program regrounding. We illustrate the framework on different optimisation problems, some of which are notoriously difficult, including shift planning and a parallel machine scheduling problem from semi-conductor production which demonstrate the effectiveness of the LNS approach.

poster session 6 @ red 2, poster session 7 @ red 2, oral session 7 @ red 2, poster session 6, poster session 7, oral session 7

The capability of a reinforcement learning (RL) agent heavily depends on the diversity of the learning scenarios generated by the environment. Generation of diverse realistic scenarios is challenging for real-time strategy (RTS) environments. The RTS environments are characterized by intelligent entities/non-RL agents cooperating and competing with the RL agents with large state and action spaces over a long period of time, resulting in an infinite space of feasible, but not necessarily realistic, scenarios involving complex interaction among different RL and non-RL agents. Yet, most of the existing simulators rely on randomly generating the environments based on predefined settings/layouts and offer limited flexibility and control over the environment dynamics for researchers to generate diverse, realistic scenarios as per their demand. To address this issue, for the first time, we formally introduce the benefits of adopting an existing formal scenario specification language, SCENIC, to assist researchers to model and generate diverse scenarios in an RTS environment in a flexible, systematic, and programmatic manner. To showcase the benefits, we interfaced SCENIC to an existing RTS environment Google Research Football (GRF) simulator and introduced a benchmark consisting of 32 realistic scenarios, encoded in SCENIC, to train RL agents and testing their generalization capabilities. We also show how researchers/RL practitioners can incorporate their domain knowledge to expedite the training process by intuitively modeling stochastic programmatic policies with SCENIC.

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

An issue that has so far received only limited attention in probabilistic logic programming (PLP) is the modelling of so-called epistemic uncertainty, the uncertainty about the model itself. Accurately quantifying this model uncertainty is paramount to robust inference, learning and ultimately decision making. We introduce BetaProbLog, a PLP language that can model epistemic uncertainty. BetaProbLog has sound semantics, an effective inference algorithm that combines Monte Carlo techniques with knowledge compilation, and a parameter learning algorithm. We emprically outperform state-of-the-art methods on probabilistic inference tasks in second-order Bayesian networks, digit classification and discriminative learning in the presence of epistemic uncertainty.

poster session 6 @ red 2, poster session 7 @ red 2, oral session 7 @ red 2, poster session 6, poster session 7, oral session 7

Participatory budgeting (PB) is a democratic process where citizens jointly decide on how to allocate public funds to indivisible projects. In this work, we focus on PB processes where citizens may provide additional money to projects they want to see funded. We introduce a formal framework for this kind of PB with donations. Our framework also allows for diversity constraints, meaning that each project belongs to one or more types, and there are lower and upper bounds on the number of projects of the same type that can be funded. We propose three general classes of methods for aggregating the citizens’ preferences in the presence of donations and analyze their axiomatic properties. Furthermore, we investigate the computational complexity of determining the outcome of a PB process with donations and of finding a citizen’s optimal donation strategy.

poster session 6 @ blue 6, poster session 10 @ blue 6, poster session 6, poster session 10

Most metric learning techniques typically focus on sample embedding learning, while implicitly assume a homogeneous local neighborhood around each sample, based on the metrics used in training (\textit{e.g.,} hypersphere for Euclidean distance or unit hyperspherical crown for cosine distance). As real-world data often lies on a low-dimensional manifold curved in a high-dimensional space, it is unlikely that everywhere of the manifold shares the same local structures in the input space. Besides, considering the non-linearity of neural networks, the local structure in the output embedding space may not be homogeneous as assumed. Therefore, representing each sample simply with its embedding while ignoring its individual neighborhood structure would have limitations in Embedding-Based Retrieval (EBR). By exploiting the heterogeneity of local structures in the embedding space, we propose a Neighborhood-Adaptive Structure Augmented metric learning framework (NASA), where the neighborhood structure is realized as a \textit{structure embedding}, and learned along with the sample embedding in a self-supervised manner. In this way, without any modifications, most indexing techniques can be used to support large-scale EBR with NASA embeddings. Experiments on six standard benchmarks with two kinds of embeddings, \textit{i.e.,} binary embeddings and real-valued embeddings, show that our method significantly improves and outperforms the state-of-the-art methods.

poster session 6 @ red 3, poster session 10 @ red 3, oral session 6 @ red 3, poster session 6, poster session 10, oral session 6

We study the problem of distributed training of neural networks (NNs) on devices with heterogeneous, limited, and time-varying availability of computational resources. We present an adaptive, resource-aware, on-device learning mechanism, DISTREAL, which is able to fully and efficiently utilize the available resources on devices in a distributed manner, increasing the convergence speed. This is achieved with a dropout mechanism that dynamically adjusts the computational complexity of training an NN by randomly dropping filters of convolutional layers of the model. Our main contribution is the introduction of a design space exploration (DSE) technique, which finds Pareto-optimal per-layer dropout vectors with respect to resource requirements and convergence speed of the training. Applying this technique, each device is able to dynamically select the dropout vector that fits its available resource without requiring any assistance from the server. We implement our solution in a federated learning (FL) system, where the availability of computational resources varies both between devices and over time, and show through extensive evaluation that we are able to significantly increase the convergence speed over the state of the art without compromising on the final accuracy.

poster session 6 @ blue 3, poster session 7 @ blue 3, poster session 6, poster session 7

Enabling computers to comprehend the intent of human actions by processing language is one of the fundamental goals of Natural Language Understanding. An emerging task in this context is that of free-form event process typing, which aims at understanding the overall goal of a protagonist in terms of an action and an object, given a sequence of events. This task was initially treated as a learning-to-rank problem by exploiting the similarity between processes and action/object textual definitions. However, this approach appears to be overly complex, binds the output types to a fixed inventory for possible word definitions and, moreover, leaves space for further enhancements as regards performance. In this paper, we advance the field by reformulating the free-form event process typing task as a sequence generation problem and put forward STEPS, an end-to-end approach for producing user intent in terms of actions and objects only, dispensing with the need for their definitions. In addition to this, we eliminate several dataset constraints set by previous works, while at the same time significantly outperforming them. We release the data and software at https://github.com/SapienzaNLP/steps.

poster session 6 @ red 5, poster session 9 @ red 5, poster session 6, poster session 9

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

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

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

Mixed-integer programming (MIP) technology offers a generic way of formulating and solving combinatorial optimization problems. While generally reliable, state-of-the-art MIP solvers base many crucial decisions on hand-crafted heuristics, largely ignoring common patterns within a given instance distribution of the problem of interest. Here, we propose MIP-GNN, a general framework for enhancing such solvers with data-driven insights. By encoding the variable-constraint interactions of a given mixed-integer linear program (MILP) as a bipartite graph, we leverage state-of-the-art graph neural network architectures to predict variable biases, i.e., component-wise averages of (near) optimal solutions, indicating how likely a variable will be set to 0 or 1 in (near) optimal solutions of binary MILPs. In turn, the predicted biases stemming from a single, once-trained model are used to guide the solver, replacing heuristic components. We integrate MIP-GNN into a state-of-the-art MIP solver, applying it to tasks such as node selection and warm-starting, showing significant improvements compared to the default setting of the solver on two classes of challenging binary MILPs.

poster session 6 @ blue 4, poster session 10 @ blue 4, oral session 10 @ blue 4, poster session 6, poster session 10, oral session 10