IJCAI 2021 论文列表

Diagnosis and treatment for Parkinson's disease rely on the evaluation of motor functions, which is expensive and time consuming when performing at clinics. It is also difficult for patients to record correct movements at home without the guidance from experienced physicians. To help patients with Parkinson’s disease get better evaluation from in-home recorded movement videos, we developed an interactive video acquisition and learning system for clinical motor assessments. The system provides real-time guidance with multi-level body keypoint tracking and analysis to patients, which guarantees correct understanding and performing of clinical tasks. We tested its effectiveness on healthy subjects, and the efficiency and usability on patient groups. Experiments showed that our system enabled high quality video recordings following clinical standards, benefiting both patients and physicians. Our system provides a novel learning-based telemedicine approach for the care of patients with Parkinson’s disease.

The ability to learn from large quantities of complex data has led to the development of intelligent agents such as self-driving cars and assistive devices. This data often comes from people via interactions such as labeling, providing rewards and punishments, and giving demonstrations or critiques. However, people's ability to provide high-quality data can be affected by human factors of an interaction, such as induced cognitive load and perceived usability. We show that these human factors differ significantly between interaction types. We first formalize interactions as a Markov Decision Process, and construct a taxonomy of these interactions to identify four archetypes: Showing, Categorizing, Sorting, and Evaluating. We then run a user study across two task domains. Our findings show that Evaluating interactions are more cognitively loading and less usable than the others, and Categorizing and Showing interactions are the least cognitively loading and most usable.

Using personalized explanations to support recommendations has been shown to increase trust and perceived quality. However, to actually obtain better recommendations, there needs to be a means for users to modify the recommendation criteria by interacting with the explanation. We present a novel technique using aspect markers that learns to generate personalized explanations of recommendations from review texts, and we show that human users significantly prefer these explanations over those produced by state-of-the-art techniques. Our work's most important innovation is that it allows users to react to a recommendation by critiquing the textual explanation: removing (symmetrically adding) certain aspects they dislike or that are no longer relevant (symmetrically that are of interest). The system updates its user model and the resulting recommendations according to the critique. This is based on a novel unsupervised critiquing method for single- and multi-step critiquing with textual explanations. Empirical results show that our system achieves good performance in adapting to the preferences expressed in multi-step critiquing and generates consistent explanations.

Reinforcement learning (RL) is a learning paradigm in which an agent interacts with the environment it inhabits to learn in a trial-and-error way. By letting the agent acquire knowledge from its own experience, RL has been successfully applied to complex domains such as robotics. However, for non-trivial problems, training an RL agent can take very long periods of time. Lifelong machine learning (LML) is a learning setting in which the agent learns to solve tasks sequentially, by leveraging knowledge accumulated from previously solved tasks to learn better/faster in a new one. Most LML works heavily rely on the assumption that tasks are similar to each other. However, this may not be true for some domains with a high degree of task-diversity that could benefit from adopting a lifelong learning approach, e.g., service robotics. Therefore, in this research we will address the problem of learning to solve a sequence of RL heterogeneous tasks (i.e., tasks that differ in their state-action space).

The use of virtual collections of data is often essential in several data and knowledge management tasks. In the literature, the standard way to define virtual data collections is via views, i.e., virtual relations defined using queries. In data and knowledge bases, the notion of views is a staple of data access, data integration and exchange, query optimization, and data privacy. In this work, we study views in Ontology-Based Data Access (OBDA) systems. OBDA is a powerful paradigm for accessing data through an ontology, i.e., a conceptual specification of the domain of interest written using logical axioms. Intuitively, users of an OBDA system interact with the data only through the ontology's conceptual lens. We present a novel framework to express natural and sophisticated forms of views in OBDA systems and introduce fundamental reasoning tasks for these views. We study the computational complexity of these tasks and present classes of views for which these tasks are tractable or at least decidable.

My research is driven by my curiosity about the nature of intelligence. Of the several aspects that characterise the behaviour of intelligent agents, I primarily study sequential decision making, learning, and exploration. My interests also extend to broader questions on the effects of AI on life and society. In this paper, I present four distinct investigations drawn from my recent work, which range from theoretical to applied, and which involve both analysis and design. I also share my outlook as an early-career researcher.

Recent researches have made remarkable achievements in fast video style transfer based on western paintings. However, due to the inherent different drawing techniques and aesthetic expressions of Chinese ink wash painting, existing methods either achieve poor temporal consistency or fail to transfer the key freehand brushstroke characteristics of Chinese ink wash painting. In this paper, we present a novel video style transfer framework for Chinese ink wash paintings. The two key ideas are a multi-frame fusion for temporal coherence and an instance-aware style transfer. The frame reordering and stylization based on reference frame fusion are proposed to improve temporal consistency. Meanwhile, the proposed method is able to adaptively leave the white spaces in the background and to select proper scales to extract features and depict the foreground subject by leveraging instance segmentation. Experimental results demonstrate the superiority of the proposed method over state-of-the-art style transfer methods in terms of both temporal coherence and visual quality. Our project website is available at https://oblivioussy.github.io/InkVideo/.

We present a review on the recent advances and emerging opportunities around the theme of analyzing deep neural networks (DNNs) with information-theoretic methods. We first discuss popular information-theoretic quantities and their estimators. We then introduce recent developments on information-theoretic learning principles (e.g., loss functions, regularizers and objectives) and their parameterization with DNNs. We finally briefly review current usages of information-theoretic concepts in a few modern machine learning problems and list a few emerging opportunities.

The selective visual attention mechanism in the human visual system (HVS) restricts the amount of information to reach visual awareness for perceiving natural scenes, allowing near real-time information processing with limited computational capacity. This kind of selectivity acts as an ‘Information Bottleneck (IB)’, which seeks a trade-off between information compression and predictive accuracy. However, such information constraints are rarely explored in the attention mechanism for deep neural networks (DNNs). In this paper, we propose an IB-inspired spatial attention module for DNN structures built for visual recognition. The module takes as input an intermediate representation of the input image, and outputs a variational 2D attention map that minimizes the mutual information (MI) between the attention-modulated representation and the input, while maximizing the MI between the attention-modulated representation and the task label. To further restrict the information bypassed by the attention map, we quantize the continuous attention scores to a set of learnable anchor values during training. Extensive experiments show that the proposed IB-inspired spatial attention mechanism can yield attention maps that neatly highlight the regions of interest while suppressing backgrounds, and bootstrap standard DNN structures for visual recognition tasks (e.g., image classification, fine-grained recognition, cross-domain classification). The attention maps are interpretable for the decision making of the DNNs as verified in the experiments. Our code is available at this https URL.

This paper introduces an algorithm for discovering implicit and delayed causal relations between events observed by a robot at regular or arbitrary times, with the objective of improving data-efficiency and interpretability of model-based reinforcement learning (RL) techniques. The proposed algorithm initially predicts observations with the Markov assumption, and incrementally introduces new hidden variables to explain and reduce the stochasticity of the observations. The hidden variables are memory units that keep track of pertinent past events. Such events are systematically identified by their information gains. A test of independence between inputs and mechanisms is performed to identify cases when there is a causal link between events and those when the information gain is due to confounding variables. The learned transition and reward models are then used in a Monte Carlo tree search for planning. Experiments on simulated and real robotic tasks, and the challenging 3D game Doom show that this method significantly improves over current RL techniques.

InfOCF-Web provides implementations of system P and system Z inference, and of inference relations based on c-representation with respect to various inference modes and different classes of minimal models. It has an easy-to-use online interface for computing ranking models of a conditional knowledge R, and for answering queries and comparing inference results of nonmonotonic inference relations induced by R.

Most of existing advantage function estimation methods in reinforcement learning suffer from the problem of high variance, which scales unfavorably with the time horizon. To address this challenge, we propose to identify the independence property between current action and future states in environments, which can be further leveraged to effectively reduce the variance of the advantage estimation. In particular, the recognized independence property can be naturally utilized to construct a novel importance sampling advantage estimator with close-to-zero variance even when the Monte-Carlo return signal yields a large variance. To further remove the risk of the high variance introduced by the new estimator, we combine it with existing Monte-Carlo estimator via a reward decomposition model learned by minimizing the estimation variance. Experiments demonstrate that our method achieves higher sample efficiency compared with existing advantage estimation methods in complex environments.

Key to the effectiveness of schedule-driven approaches to real-time traffic control is an ability to accurately predict when sensed vehicles will arrive at and pass through the intersection. Prior work in schedule-driven traffic control has assumed a static vehicle arrival model. However, this static predictive model ignores the fact that the queue count and the incurred delay should vary as different partial signal timing schedules (i.e., different possible futures) are explored during the online planning process. In this paper, we propose an alternative arrival time model that incorporates queueing dynamics into this forward search process for a signal timing schedule, to more accurately capture how the intersection’s queues vary over time. As each search state is generated, an incremental queueing delay is dynamically projected for each vehicle. The resulting total queueing delay is then considered in addition to the cumulative delay caused by signal operations. We demonstrate the potential of this approach through microscopic traffic simulation of a real-world road network, showing a 10-15% reduction in average wait times over the schedule-driven traffic signal control system in heavy traffic scenarios.

One of the main aims of the methods developed for reasoning under inconsistency, in particular paraconsistent inference, is to derive informative conclusions from inconsistent bases. In this paper, we introduce an approach based on inconsistency measurement for defining non-monotonic paraconsistent consequence relations. The main idea consists in adapting properties of classical reasoning under consistency to inconsistent propositional bases by involving inconsistency measures (IM). We first exhibit interesting properties of our consequence relations. We then study situations where they bring about consequences that are always jointly consistent. In particular, we introduce a property of inconsistency measures that guarantees the consistency of the set of all entailed formulas. We also show that this property leads to several interesting properties of our IM-based consequence relations. Finally, we discuss relationships between our framework and well-known consequence relations that are based on maximal consistent subsets. In this setting, we establish direct connections between the latter and properties of inconsistency measures.

We study one-sided matching problems where each agent must be assigned at most one object. In this classic problem it is often assumed that agents specify only ordinal preferences over objects and the goal is to return a matching that satisfies some desirable property such as Pareto optimality or rank-maximality. However, agents may have cardinal utilities describing their preference intensities and ignoring this can result in welfare loss. We investigate how to elicit additional cardinal information from agents using simple threshold queries and use it in turn to design algorithms that return a matching satisfying a desirable matching property, while also achieving a good approximation to the optimal welfare among all matchings satisfying that property. Overall, our results show how one can improve welfare by even non-adaptively asking agents for just one bit of extra information per object.

Due to different types of inputs, diverse text generation tasks may adopt different encoder-decoder frameworks. Thus most existing approaches that aim to improve the robustness of certain generation tasks are input-relevant, and may not work well for other generation tasks. Alternatively, in this paper we present a universal approach to enhance the language representation for text generation on the base of generic encoder-decoder frameworks. This is done from two levels. First, we introduce randomness by randomly masking some percentage of tokens on the decoder side when training the models. In this way, instead of using ground truth history context, we use its corrupted version to predict the next token. Then we propose an auxiliary task to properly recover those masked tokens. Experimental results on several text generation tasks including machine translation (MT), AMR-to-text generation, and image captioning show that the proposed approach can significantly improve over competitive baselines without using any task-specific techniques. This suggests the effectiveness and generality of our proposed approach.

Stylized neural machine translation (NMT) aims to translate sentences of one style into sentences of another style, which is essential for the application of machine translation in a real-world scenario. However, a major challenge in this task is the scarcity of high-quality parallel data which is stylized paired. To address this problem, we propose an iterative dual knowledge transfer framework that utilizes informal training data of machine translation and formality style transfer data to create large-scale stylized paired data, for the training of stylized machine translation model. Specifically, we perform bidirectional knowledge transfer between translation model and text style transfer model iteratively through knowledge distillation. Then, we further propose a data-refinement module to process the noisy synthetic parallel data generated during knowledge transfer. Experiment results demonstrate the effectiveness of our method, achieving an improvement over the existing best model by 5 BLEU points on MTFC dataset. Meanwhile, extensive analyses illustrate our method can also improve the accuracy of formality style transfer.

Most sequential recommendation models capture the features of consecutive items in a user-item interaction history. Though effective, their representation expressiveness is still hindered by the sparse learning signals. As a result, the sequential recommender is prone to make inconsistent predictions. In this paper, we propose a model, SSI, to improve sequential recommendation consistency with Self-Supervised Imitation. Precisely, we extract the consistency knowledge by utilizing three self-supervised pre-training tasks, where temporal consistency and persona consistency capture user-interaction dynamics in terms of the chronological order and persona sensitivities, respectively. Furthermore, to provide the model with a global perspective, global session consistency is introduced by maximizing the mutual information among global and local interaction sequences. Finally, to comprehensively take advantage of all three independent aspects of consistency-enhanced knowledge, we establish an integrated imitation learning framework. The consistency knowledge is effectively internalized and transferred to the student model by imitating the conventional prediction logit as well as the consistency-enhanced item representations. In addition, the flexible self-supervised imitation framework can also benefit other student recommenders. Experiments on four real-world datasets show that SSI effectively outperforms the state-of-the-art sequential recommendation methods.

We present a multi-agent learning algorithm, ALMA-Learning, for efficient and fair allocations in large-scale systems. We circumvent the traditional pitfalls of multi-agent learning (e.g., the moving target problem, the curse of dimensionality, or the need for mutually consistent actions) by relying on the ALMA heuristic as a coordination mechanism for each stage game. ALMA-Learning is decentralized, observes only own action/reward pairs, requires no inter-agent communication, and achieves near-optimal (<5% loss) and fair coordination in a variety of synthetic scenarios and a real-world meeting scheduling problem. The lightweight nature and fast learning constitute ALMA-Learning ideal for on-device deployment.