IJCAI 2020 论文列表

Tandem mass spectrometry is the most widely used technology to identify proteins in a complex biological sample, which produces a large number of spectra representative of protein subsequences named peptide. In this paper, we propose a hierarchical multi-stage framework, referred as DeepTag, to identify the peptide sequence for each given spectrum. Compared with the traditional one-stage generation, our sequencing model starts the inference with a selected high-confidence guiding tag and provides the complete sequence based on this guiding tag. Besides, we introduce a cross-modality refining module to asist the decoder focus on effective peaks and fine-tune with a reinforcement learning technique. Experiments on different public datasets demonstrate that our method achieves a new state-of-the-art performance in peptide identification task, leading to a marked improvement in terms of both precision and recall.

Knowledge tracing (KT) defines the task of predicting whether students can correctly answer questions based on their historical response. Although much research has been devoted to exploiting the question information, plentiful advanced information among questions and skills hasn't been well extracted, making it challenging for previous work to perform adequately. In this paper, we demonstrate that large gains on KT can be realized by pre-training embeddings for each question on abundant side information, followed by training deep KT models on the obtained embeddings. To be specific, the side information includes question difficulty and three kinds of relations contained in a bipartite graph between questions and skills. To pre-train the question embeddings, we propose to use product-based neural networks to recover the side information. As a result, adopting the pre-trained embeddings in existing deep KT models significantly outperforms state-of-the-art baselines on three common KT datasets.

Graph Neural Networks (GNNs) are powerful for the representation learning of graph-structured data. Most of the GNNs use a message-passing scheme, where the embedding of a node is iteratively updated by aggregating the information from its neighbors. To achieve a better expressive capability of node influences, attention mechanism has grown to be popular to assign trainable weights to the nodes in aggregation. Though the attention-based GNNs have achieved remarkable results in various tasks, a clear understanding of their discriminative capacities is missing. In this work, we present a theoretical analysis of the representational properties of the GNN that adopts the attention mechanism as an aggregator. Our analysis determines all cases when those attention-based GNNs can always fail to distinguish certain distinct structures. Those cases appear due to the ignorance of cardinality information in attention-based aggregation. To improve the performance of attention-based GNNs, we propose cardinality preserved attention (CPA) models that can be applied to any kind of attention mechanisms. Our experiments on node and graph classification confirm our theoretical analysis and show the competitive performance of our CPA models. The code is available online: https://github.com/zetayue/CPA.

Previous research has claimed dynamic epistemic logic (DEL) to be a suitable formalism for representing essential aspects of a Theory of Mind (ToM) for an autonomous agent. This includes the ability of the formalism to represent the reasoning involved in false-belief tasks of arbitrary order, and hence for autonomous agents based on the formalism to become able to pass such tests. This paper provides evidence for the claims by documenting the implementation of a DEL-based reasoning system on a humanoid robot. Our implementation allows the robot to perform cognitive perspective-taking, in particular to reason about the first- and higher-order beliefs of other agents. We demonstrate how this allows the robot to pass a quite general class of false-belief tasks involving human agents. Additionally, as is briefly illustrated, it allows the robot to proactively provide human agents with relevant information in situations where a system without ToM-abilities would fail. The symbolic grounding problem of turning robotic sensor input into logical action descriptions in DEL is achieved via a perception system based on deep neural networks.

We propose a novel framework to identify sub-goals useful for exploration in sequential decision making tasks under partial observability. We utilize the variational intrinsic control framework (Gregor et.al., 2016) which maximizes empowerment -- the ability to reliably reach a diverse set of states and show how to identify sub-goals as states with high necessary option information through an information theoretic regularizer. Despite being discovered without explicit goal supervision, our sub-goals provide better exploration and sample complexity on challenging grid-world navigation tasks compared to supervised counterparts in prior work.

Inverse kinematics solves the problem of how to control robot arm joints to achieve desired end effector positions, which is critical to any robot arm design and implementations of control algorithms. It is a common misunderstanding that closed-form inverse kinematics analysis is solved. Popular software and algorithms, such as gradient descent or any multi-variant equations solving algorithm, claims solving inverse kinematics but only on the numerical level. While the numerical inverse kinematics solutions are relatively straightforward to obtain, these methods often fail, even when the inverse kinematics solutions exist. Therefore, closed-form inverse kinematics analysis is superior, but there is no generalized automated algorithm. Up till now, the high-level logical reasoning involved in solving closed-form inverse kinematics made it hard to automate, so it's handled by human experts. We developed IKBT, a knowledge-based intelligent system that can mimic human experts' behaviors in solving closed-from inverse kinematics using Behavior Tree. Knowledge and rules used by engineers when solving closed-from inverse kinematics are encoded as actions in Behavior Tree. The order of applying these rules is governed by higher level composite nodes, which resembles the logical reasoning process of engineers. It is also the first time that the dependency of joint variables, an important issue in inverse kinematics analysis, is automatically tracked in graph form. Besides generating closed-form solutions, IKBT also explains its solving strategies in human (engineers) interpretable form. This is a proof-of-concept of using Behavior Trees to solve high-cognitive problems.

Networked observational data presents new opportunities for learning individual causal effects, which plays an indispensable role in decision making. Such data poses the challenge of confounding bias. Previous work presents two desiderata to handle confounding bias. On the treatment group level, we aim to balance the distributions of confounder representations. On the individual level, it is desirable to capture patterns of hidden confounders that predict treatment assignments. Existing methods show the potential of utilizing network information to handle confounding bias, but they only try to satisfy one of the two desiderata. This is because the two desiderata seem to contradict each other. When the two distributions of confounder representations are highly overlapped, then we confront the undiscriminating problem between the treated and the controlled. In this work, we formulate the two desiderata as a minimax game. We propose IGNITE that learns representations of confounders from networked observational data, which is trained by a minimax game to achieve the two desiderata. Experiments verify the efficacy of IGNITE on two datasets under various settings.

Learning expressive representations is always crucial for well-performed policies in deep reinforcement learning (DRL). Different from supervised learning, in DRL, accurate targets are not always available, and some inputs with different actions only have tiny differences, which stimulates the demand for learning expressive representations. In this paper, firstly, we empirically compare the representations of DRL models with different performances. We observe that the representations of a better state extractor (SE) are more scattered than a worse one when they are visualized. Thus, we investigate the singular values of representation matrix, and find that, better SEs always correspond to smaller differences among these singular values. Next, based on such observations, we define an indicator of the representations for DRL model, which is the Number of Significant Singular Values (NSSV) of a representation matrix. Then, we propose I4R algorithm, to improve DRL algorithms by adding the corresponding regularization term to enhance the NSSV. Finally, we apply I4R to both policy gradient and value based algorithms on Atari games, and the results show the superiority of our proposed method.

Online kernel selection in continuous kernel space is more complex than that in discrete kernel set. But existing online kernel selection approaches for continuous kernel spaces have linear computational complexities at each round with respect to the current number of rounds and lack sublinear regret guarantees due to the continuously many candidate kernels. To address these issues, we propose a novel hypothesis sketching approach to online kernel selection in continuous kernel space, which has constant computational complexities at each round and enjoys a sublinear regret bound. The main idea of the proposed hypothesis sketching approach is to maintain the orthogonality of the basis functions and the prediction accuracy of the hypothesis sketches in a time-varying reproducing kernel Hilbert space. We first present an efficient dependency condition to maintain the basis functions of the hypothesis sketches under a computational budget. Then we update the weights and the optimal kernels by minimizing the instantaneous loss of the hypothesis sketches using the online gradient descent with a compensation strategy. We prove that the proposed hypothesis sketching approach enjoys a regret bound of order O(√T) for online kernel selection in continuous kernel space, which is optimal for convex loss functions, where T is the number of rounds, and reduces the computational complexities at each round from linear to constant with respect to the number of rounds. Experimental results demonstrate that the proposed hypothesis sketching approach significantly improves the efficiency of online kernel selection in continuous kernel space while retaining comparable predictive accuracies.

Personalized news recommendation can help users stay on top of the current affairs without being overwhelmed by the endless torrents of online news. However, the freshness or timeliness of news has been largely ignored by current news recommendation systems. In this paper, we propose a novel approach dubbed HyperNews which explicitly models the effect of timeliness on news recommendation. Furthermore, we introduce an auxiliary task of predicting the so-called "active-time" that users spend on each news article. Our key finding is that it is beneficial to address the problem of news recommendation together with the related problem of active-time prediction in a multi-task learning framework. Specifically, we train a double-task deep neural network (with a built-in timeliness module) to carry out news recommendation and active-time prediction simultaneously. To the best of our knowledge, such a "kill-two-birds-with-one-stone" solution has seldom been tried in the field of news recommendation before. Our extensive experiments on real-life news datasets have not only confirmed the mutual reinforcement of news recommendation and active-time prediction but also demonstrated significant performance improvements over state-of-the-art news recommendation techniques.

Hyperbolic space is a well-defined space with constant negative curvature. Recent research demonstrates its odds of capturing complex hierarchical structures with its exceptional high capacity and continuous tree-like properties. This paper bridges hyperbolic space's superiority to the power-law structure of documents by introducing a hyperbolic neural network architecture named Hyperbolic Hierarchical Attention Network (Hype-HAN). Hype-HAN defines three levels of embeddings (word/sentence/document) and two layers of hyperbolic attention mechanism (word-to-sentence/sentence-to-document) on Riemannian geometries of the Lorentz model, Klein model and Poincaré model. Situated on the evolving embedding spaces, we utilize both conventional GRUs (Gated Recurrent Units) and hyperbolic GRUs with Möbius operations. Hype-HAN is applied to large scale datasets. The empirical experiments show the effectiveness of our method.

This paper aims to learn multi-agent cooperation where each agent performs its actions in a decentralized way. In this case, it is very challenging to learn decentralized policies when the rewards are global and sparse. Recently, learning from demonstrations (LfD) provides a promising way to handle this challenge. However, in many practical tasks, the available demonstrations are often sub-optimal. To learn better policies from these sub-optimal demonstrations, this paper follows a centralized learning and decentralized execution framework and proposes a novel hybrid learning method based on multi-agent actor-critic. At first, the expert trajectory returns generated from demonstration actions are used to pre-train the centralized critic network. Then, multi-agent decisions are made by best response dynamics based on the critic and used to train the decentralized actor networks. Finally, the demonstrations are updated by the actor networks, and the critic and actor networks are learned jointly by running the above two steps alliteratively. We evaluate the proposed approach on a real-time strategy combat game. Experimental results show that the approach outperforms many competing demonstration-based methods.

Deep neural networks are usually considered black-boxes due to their complex internal architecture, that cannot straightforwardly provide human-understandable explanations on how they behave. Indeed, Deep Learning is still viewed with skepticism in those real-world domains in which incorrect predictions may produce critical effects. This is one of the reasons why in the last few years Explainable Artificial Intelligence (XAI) techniques have gained a lot of attention in the scientific community. In this paper, we focus on the case of multi-label classification, proposing a neural network that learns the relationships among the predictors associated to each class, yielding First-Order Logic (FOL)-based descriptions. Both the explanation-related network and the classification-related network are jointly learned, thus implicitly introducing a latent dependency between the development of the explanation mechanism and the development of the classifiers. Our model can integrate human-driven preferences that guide the learning-to-explain process, and it is presented in a unified framework. Different typologies of explanations are evaluated in distinct experiments, showing that the proposed approach discovers new knowledge and can improve the classifier performance.

In this study, we present a unique demographically representative dataset of 15k US residents that combines technology use logs with surveys on moral views, human values, and emotional contagion. First, we show which values determine the adoption of Health & Fitness mobile applications, finding that users who prioritize the value of purity and de-emphasize values of conformity, hedonism, and security are more likely to use such apps. Further, we achieve a weighted AUROC of .673 in predicting whether individual exercises and find a strong link of exercise to respondent socioeconomic status, as well as the value of loyalty.

Human gaze reveals a wealth of information about internal cognitive state. Thus, gaze-related research has significantly increased in computer vision, natural language processing, decision learning, and robotics in recent years. We provide a high-level overview of the research efforts in these fields, including collecting human gaze data sets, modeling gaze behaviors, and utilizing gaze information in various applications, with the goal of enhancing communication between these research areas. We discuss future challenges and potential applications that work towards a common goal of human-centered artificial intelligence.

Image captioning aims to describe an image with a concise, accurate, and interesting sentence. To build such an automatic neural captioner, the traditional models align the generated words with a number of human-annotated sentences to mimic human-like captions. However, the crowd-sourced annotations inevitably come with data quality issues such as grammatical errors, wrong identification of visual objects and sub-optimal sentence focus. During the model training, existing methods treat all the annotations equally regardless of the data quality. In this work, we explicitly engage human consensus to measure the quality of ground truth captions in advance, and directly encourage the model to learn high quality captions with high priority. Therefore, the proposed consensus-oriented method can accelerate the training process and achieve superior performance with only supervised objective without time-consuming reinforcement learning. The novel consensus loss can be implemented into most of the existing state-of-the-art methods, boosting the BLEU-4 performance by maximum relative 12.47% comparing to the conventional cross-entropy loss. Extensive experiments are conducted on MS-COCO Image Captioning dataset demonstrating the proposed human consensus-oriented training method can significantly improve the training efficiency and model effectiveness.

Recently semantic parsing in context has received a considerable attention, which is challenging since there are complex contextual phenomena. Previous works verified their proposed methods in limited scenarios, which motivates us to conduct an exploratory study on context modeling methods under real-world semantic parsing in context. We present a grammar-based decoding semantic parser and adapt typical context modeling methods on top of it. We evaluate 13 context modeling methods on two large complex cross-domain datasets, and our best model achieves state-of-the-art performances on both datasets with significant improvements. Furthermore, we summarize the most frequent contextual phenomena, with a fine-grained analysis on representative models, which may shed light on potential research directions. Our code is available at https://github.com/microsoft/ContextualSP.

The efficiency of modern automotive body shop assembly lines is highly related to the reduction of downtimes due to failures and quality deviations within the manufacturing process. Consequently, the need for implementing tools into the assembly lines for on-line monitoring, and failure diagnosis, also under the prism of improving the troubleshooting, is of great importance. While the identification of root causes and elimination of failures is usually built upon individual on-site expert knowledge, causal graphical models (CGMs) have opened the possibility to make a purely data-driven assessment. In this demo, we showcase how a CGM of the production process is incorporated into a monitoring tool to function as a decision-support system for an operator of a modern automotive body shop assembly line and enables fast and effective handling of failures and quality deviations.

In interactive e-learning environments such as Intelligent Tutoring Systems, there are pedagogical decisions to make at two main levels of granularity: whole problems and single steps. In recent years, there is growing interest in applying data-driven techniques for adaptive decision making that can dynamically tailor students' learning experiences. Most existing data-driven approaches, however, treat these pedagogical decisions equally, or independently, disregarding the long-term impact that tutor decisions may have across these two levels of granularity. In this paper, we propose and apply an offline Gaussian Processes based Hierarchical Reinforcement Learning (HRL) framework to induce a hierarchical pedagogical policy that makes decisions at both problem and step levels. An empirical classroom study shows that the HRL policy is significantly more effective than a Deep Q-Network (DQN) induced policy and a random yet reasonable baseline policy.