AAAI 2021 Papers

We propose a model of winners allocation. In this model, we are given are two elections where the sets of candidates may intersect. The goal is to find two disjoint winning committees from respectively the two elections that are subjected to certain reasonable restrictions. For our model, we first propose several desirable properties. Then, we investigate the implication relationships among these properties. Finally, we study the complexity of computing winners allocations providing these properties. For hardness results, we also study some fixed-parameter algorithms.

We propose a market-inspired bidding scheme for the assignment of paper reviews in large academic conferences. We provide an analysis of the incentives of reviewers during the bidding phase, when reviewers have both private costs and some information about the demand for each paper; and their goal is to obtain the best possible k papers for a predetermined k. We show that by assigning `budgets' to reviewers and a `price' for every paper that is (roughly) proportional to its demand, the best response of a reviewer is to bid sincerely, i.e., on her most favorite papers, and match the budget even when it is not enforced. This game-theoretic analysis is based on a simple, prototypical assignment algorithm. We show via extensive simulations on bidding data from real conferences, that our bidding scheme would substantially improve both the bid distribution and the resulting assignment.

Biomedical entity linking aims to map biomedical mentions, such as diseases and drugs, to standard entities in a given knowledge base. The specific challenge in this context is that the same biomedical entity can have a wide range of names, including synonyms, morphological variations, and names with different word orderings. Recently, BERT-based methods have advanced the state-of-the-art by allowing for rich representations of word sequences. However, they often have hundreds of millions of parameters and require heavy computing resources, which limits their applications in resource-limited scenarios. Here, we propose a lightweight neural method for biomedical entity linking, which needs just a fraction of the parameters of a BERT model and much less computing resources. Our method uses a simple alignment layer with attention mechanisms to capture the variations between mention and entity names. Yet, we show that our model is competitive with previous work on standard evaluation benchmarks.

Aspect based sentiment analysis (ABSA) involves three fundamental subtasks: aspect term extraction, opinion term extraction, and aspect-level sentiment classification. Early works only focused on solving one of these subtasks individually. Some recent work focused on solving a combination of two subtasks, e.g., extracting aspect terms along with sentiment polarities or extracting the aspect and opinion terms pair-wisely. More recently, the triple extraction task has been proposed, i.e., extracting the (aspect term, opinion term, sentiment polarity) triples from a sentence. However, previous approaches fail to solve all subtasks in a unified end-to-end framework. In this paper, we propose a complete solution for ABSA. We construct two machine reading comprehension (MRC) problems, and solve all subtasks by joint training two BERT-MRC models with parameters sharing. We conduct experiments on these subtasks and results on several benchmark datasets demonstrate the effectiveness of our proposed framework, which significantly outperforms existing state-of-the-art methods.

Recent theoretical analyses reveal that existing Stochastic Gradient Markov Chain Monte Carlo (SG-MCMC) methods need large mini-batches of samples (exponentially dependent on the dimension) to reduce the mean square error of gradient estimates and ensure non-asymptotic convergence guarantees when the target distribution has a nonconvex potential function. In this paper, we propose a novel SG-MCMC algorithm, called Hybrid Stochastic Gradient Hamiltonian Monte Carlo (HSG-HMC) method, which needs merely one sample per iteration and possesses a simple structure with only one hyperparameter. Such improvement leverages a hybrid stochastic gradient estimator that exploits historical stochastic gradient information to control the mean square error. Theoretical analyses show that our method obtains the best-known overall sample complexity to achieve epsilon-accuracy in terms of the 2-Wasserstein distance for sampling from distributions with nonconvex potential functions. Empirical studies on both simulated and real-world datasets demonstrate the advantage of our method.

Tables of data are used to record vast amounts of socioeconomic, scientific, and governmental information. Although humans create tables using underlying organizational principles, unfortunately AI systems struggle to understand the contents of these tables. This paper introduces an end-to-end system for table understanding, the process of capturing the relational structure of data in tables. We introduce models that identify cell types, group these cells into blocks of data that serve a similar functional role, and predict the relationships between these blocks. We introduce a hybrid, neuro-symbolic approach, combining embedded representations learned from thousands of tables with probabilistic constraints that capture regularities in how humans organize tables. Our neuro-symbolic model is better able to capture positional invariants of headers and enforce homogeneity of data types. One limitation in this research area is the lack of rich datasets for evaluating end-to-end table understanding, so we introduce a new benchmark dataset comprised of 431 diverse tables from data.gov. The evaluation results show that our system achieves the state-of-the-art performance on cell type classification, block identification, and relationship prediction, improving over prior efforts by up to 7% of macro F1 score.

The interactive recommender systems involve users in the recommendation procedure by receiving timely user feedback to update the recommendation policy. Therefore, they are widely used in real application scenarios. Previous interactive recommendation methods primarily focus on learning users' personalized preferences on the relevance properties of an item set. However, the investigation of users' personalized preferences on the diversity properties of an item set is usually ignored. To overcome this problem, we propose the Linear Modular Dispersion Bandit (LMDB) framework, which is an online learning setting for optimizing a combination of modular functions and dispersion functions. Specifically, LMDB employs modular functions to model the relevance properties of each item, and dispersion functions to describe the diversity properties of an item set. Moreover, we also develop a learning algorithm, called Linear Modular Dispersion Hybrid (LMDH) to solve the LMDB problem and derive a gap-free bound on its n-step regret. Extensive experiments on real datasets are performed to demonstrate the effectiveness of the proposed LMDB framework in balancing the recommendation accuracy and diversity.

Weakly supervised temporal action localization is a challenging vision task due to the absence of ground-truth temporal locations of actions in the training videos. With only video-level supervision during training, most existing methods rely on a Multiple Instance Learning (MIL) framework to predict the start and end frame of each action category in a video. However, the existing MIL-based approach has a major limitation of only capturing the most discriminative frames of an action, ignoring the full extent of an activity. Moreover, these methods cannot model background activity effectively, which plays an important role in localizing foreground activities. In this paper, we present a novel framework named HAM-Net with a hybrid attention mechanism which includes temporal soft, semi-soft and hard attentions to address these issues. Our temporal soft attention module, guided by an auxiliary background class in the classification module, models the background activity by introducing an ``action-ness'' score for each video snippet. Moreover, our temporal semi-soft and hard attention modules, calculating two attention scores for each video snippet, help to focus on the less discriminative frames of an action to capture the full action boundary. Our proposed approach outperforms recent state-of-the-art methods by at least 2.2% mAP at IoU threshold 0.5 on the THUMOS14 dataset, and by at least 1.3% mAP at IoU threshold 0.75 on the ActivityNet1.2 dataset.

In multi-event survival analysis, one aims to predict the probability of multiple different events occurring over some time horizon. One typically assumes that the timing of events is drawn from some distribution conditioned on an individual's covariates. However, during training, one does not have access to this distribution, and the natural variation in the observed event times makes the task of survival prediction challenging, on top of the potential interdependence among events. To address this issue, we introduce a novel approach for multi-event survival analysis that models the probability of event occurrence hierarchically at different time scales, using coarse predictions (e.g., monthly predictions) to iteratively guide predictions at finer and finer grained time scales (e.g., daily predictions). We evaluate the proposed approach across several publicly available datasets in terms of both intra-event, inter-individual (global) and intra-individual, inter-event (local) consistency. We show that the proposed method consistently outperforms well-accepted and commonly used approaches to multi-event survival analysis. When estimating survival curves for Alzheimer's disease and mortality, our approach achieves a C-index of 0.91 (95% CI 0.88-0.93) and a local consistency score of 0.97 (95% CI 0.94-0.98) compared to a C-index of 0.75 (95% CI 0.70-0.80) and a local consistency score of 0.94 (95% CI 0.91-0.97) when modeling each event separately. Overall, our approach improves the accuracy of survival predictions by iteratively reducing the original task to a set of nested, simpler subtasks.

To retrieve more relevant, appropriate and useful documents given a query, finding clues about that query through the text is crucial. Recent deep learning models regard the task as a term-level matching problem, which seeks exact or similar query patterns in the document. However, we argue that they are inherently based on local interactions and do not generalise to ubiquitous, non-consecutive contextual relationships. In this work, we propose a novel relevance matching model based on graph neural networks to leverage the document-level word relationships for ad-hoc retrieval. In addition to the local interactions, we explicitly incorporate all contexts of a term through the graph-of-word text format. Matching patterns can be revealed accordingly to provide a more accurate relevance score. Our approach significantly outperforms strong baselines on two ad-hoc benchmarks. We also experimentally compare our model with BERT and show our advantages on long documents.

We investigate response selection for multi-turn conversation in retrieval-based chatbots. Existing studies pay more attention to the matching between utterances and responses by calculating the matching score based on learned features, leading to insufficient model reasoning ability. In this paper, we propose a graph- reasoning network (GRN) to address the problem. GRN first conducts pre-training based on ALBERT using next utterance prediction and utterance order prediction tasks specifically devised for response selection. These two customized pre-training tasks can endow our model with the ability of capturing semantical and chronological dependency between utterances. We then fine-tune the model on an integrated network with sequence reasoning and graph reasoning structures. The sequence reasoning module conducts inference based on the highly summarized context vector of utterance-response pairs from the global perspective. The graph reasoning module conducts the reasoning on the utterance-level graph neural network from the local perspective. Experiments on two conversational reasoning datasets show that our model can dramatically outperform the strong baseline methods and can achieve performance which is close to human-level.

Self-Supervised monocular visual odometry (VO) is often cast into a view synthesis problem based on depth and camera pose estimation. One of the key challenges is to accurately and robustly estimate depth with occlusions and moving objects in the scene. Existing methods simply detect and mask out regions of occlusions locally by several convolutional layers, and then perform only partial view synthesis in the rest of the image. However, occlusion and moving object detection is an unsolved problem itself which requires global layout information. Inaccurate detection inevitably results in incorrect depth as well as pose estimation. In this work, instead of locally detecting and masking out occlusions and moving objects, we propose to alleviate their negative effects on monocular VO implicitly but more effectively from two global perspectives. First, a multi-scale non-local attention module, consisting of both intra-stage augmented attention and cascaded across-stage attention, is proposed for robust depth estimation given occlusions, alleviating the impacts of occlusions via global attention modeling. Second, adversarial learning is introduced in view synthesis for monocular VO. Unlike existing methods that use pixel-level losses on the quality of synthesized views, we enforce the synthetic view to be indistinguishable from the real one at the scene-level. Such a global constraint again helps cope with occluded and moving regions. Extensive experiments on the KITTI dataset show that our approach achieves new state-of-the-art in both pose estimation and depth recovery.

Causal inference from observational data is receiving wide applications in many fields. However, unidentifiable situations, where causal effects cannot be uniquely computed from observational data, pose critical barriers to applying causal inference to complicated real applications. In this paper, we develop a bounding method for estimating the average causal effect (ACE) under unidentifiable situations due to hidden confounding based on Pearl's structural causal model. We propose to parameterize the unknown exogenous random variables and structural equations of a causal model using neural networks and implicit generative models. Then, using an adversarial learning framework, we search the parameter space to explicitly traverse causal models that agree with the given observational distribution, and find those that minimize or maximize the ACE to obtain its lower and upper bounds. The proposed method does not make assumption about the type of structural equations and variables. Experiments using both synthetic and real-world datasets are conducted.

The paper discusses theoretical foundations that describe principles and processes involved in defining semantics that deal with similarity between arguments. Such semantics compute the strength of an argument on the basis of the strengths of its attackers, similarities between those attackers, and an initial weight ascribed to the argument. We define a semantics by three functions: an adjustment function that updates the strengths of attackers on the basis of their similarities, an aggregation function that computes the strength of the group of attackers, and an influence function that evaluates the impact of the group on the argument's initial weight. We propose intuitive constraints for the three functions and key rationality principles for semantics, and show how the former lead to the satisfaction of the latter. Then, we propose a broad family of semantics whose instances satisfy the principles. Finally, we analyse the existing adjustment functions and show that they violate some properties, then we propose novel ones and use them for generalizing h-Categorizer.

This paper studies the problem of learning interactive recommender systems from logged feedbacks without any exploration in online environments. We address the problem by proposing a general offline reinforcement learning framework for recommendation, which enables maximizing cumulative user rewards without online exploration. Specifically, we first introduce a probabilistic generative model for interactive recommendation, and then propose an effective inference algorithm for discrete and stochastic policy learning based on logged feedbacks. In order to perform offline learning more effectively, we propose five approaches to minimize the distribution mismatch between the logging policy and recommendation policy: support constraints, supervised regularization, policy constraints, dual constraints and reward extrapolation. We conduct extensive experiments on two public real-world datasets, demonstrating that the proposed methods can achieve superior performance over existing supervised learning and reinforcement learning methods for recommendation.

We propose a novel framework that unifies and extends existing methods of transfer learning (TL) for regression. To bridge a pretrained source model to the model on a target task, we introduce a density-ratio reweighting function, which is estimated through the Bayesian framework with a specific prior distribution. By changing two intrinsic hyperparameters and the choice of the density-ratio model, the proposed method can integrate three popular methods of TL: TL based on cross-domain similarity regularization, a probabilistic TL using the density-ratio estimation, and fine-tuning of pretrained neural networks. Moreover, the proposed method can benefit from its simple implementation without any additional cost; the regression model can be fully trained using off-the-shelf libraries for supervised learning in which the original output variable is simply transformed to a new output variable. We demonstrate its simplicity, generality, and applicability using various real data applications.

Unsupervised domain adaptation (UDA) assumes that source and target domain data are freely available and usually trained together to reduce the domain gap. However, considering the data privacy and the inefficiency of data transmission, it is impractical in real scenarios. Hence, it draws our eyes to optimize the network in the target domain without accessing labeled source data. To explore this direction in object detection, for the first time, we propose a source data-free domain adaptive object detection (SFOD) framework via modeling it into a problem of learning with noisy labels. Generally, a straightforward method is to leverage the pre-trained network from the source domain to generate the pseudo labels for target domain optimization. However, it is difficult to evaluate the quality of pseudo labels since no labels are available in target domain. In this paper, self-entropy descent (SED) is a metric proposed to search an appropriate confidence threshold for reliable pseudo label generation without using any handcrafted labels. Nonetheless, completely clean labels are still unattainable. After a thorough experimental analysis, false negatives are found to dominate in the generated noisy labels. Undoubtedly, false negatives mining is helpful for performance improvement, and we ease it to false negatives simulation through data augmentation like Mosaic. Extensive experiments conducted in four representative adaptation tasks have demonstrated that the proposed framework can easily achieve state-of-the-art performance. From another view, it also reminds the UDA community that the labeled source data are not fully exploited in the existing methods.

With the increasing scale of machine learning tasks, it has become essential to reduce the communication between computing nodes. Early work on gradient compression focused on the bottleneck between CPUs and GPUs, but communication-efficiency is now needed in a variety of different system architectures, from high-performance clusters to energy-constrained IoT devices. In the current practice, compression levels are typically chosen before training and settings that work well for one task may be vastly suboptimal for another dataset on another architecture. In this paper, we propose a flexible framework which adapts the compression level to the true gradient at each iteration, maximizing the improvement in the objective function that is achieved per communicated bit. Our framework is easy to adapt from one technology to the next by modeling how the communication cost depends on the compression level for the specific technology. Theoretical results and practical experiments indicate that the automatic tuning strategies significantly increase communication efficiency on several state-of-the-art compression schemes.

We consider the one-sided matching problem, where n agents have preferences over n items, and these preferences are induced by underlying cardinal valuation functions. The goal is to match every agent to a single item so as to maximize the social welfare. Most of the related literature, however, assumes that the values of the agents are not a priori known, and only access to the ordinal preferences of the agents over the items is provided. Consequently, this incomplete information leads to loss of efficiency, which is measured by the notion of distortion. In this paper, we further assume that the agents can answer a small number of queries, allowing us partial access to their values. We study the interplay between elicited cardinal information (measured by the number of queries per agent) and distortion for one-sided matching, as well as a wide range of well-studied related problems. Qualitatively, our results show that with a limited number of queries, it is possible to obtain significant improvements over the classic setting, where only access to ordinal information is given.