IJCAI 2021 论文列表

Video data is multimodal in its nature, where an utterance can involve linguistic, visual and acoustic information. Therefore, a key challenge for video sentiment analysis is how to combine different modalities for sentiment recognition effectively. The latest neural network approaches achieve state-of-the-art performance, but they neglect to a large degree of how humans understand and reason about sentiment states. By contrast, recent advances in quantum probabilistic neural models have achieved comparable performance to the state-of-the-art, yet with better transparency and increased level of interpretability. However, the existing quantum-inspired models treat quantum states as either a classical mixture or as a separable tensor product across modalities, without triggering their interactions in a way that they are correlated or non-separable (i.e., entangled). This means that the current models have not fully exploited the expressive power of quantum probabilities. To fill this gap, we propose a transparent quantum probabilistic neural model. The model induces different modalities to interact in such a way that they may not be separable, encoding crossmodal information in the form of non-classical correlations. Comprehensive evaluation on two benchmarking datasets for video sentiment analysis shows that the model achieves significant performance improvement. We also show that the degree of non-separability between modalities optimizes the post-hoc interpretability.

In recent years, the axiomatic approach to centrality measures has attracted attention in the literature. However, most papers propose a collection of axioms dedicated to one or two considered centrality measures. In result, it is hard to capture the differences and similarities between various measures. In this paper, we propose an axiom system for four classic feedback centralities: Eigenvector centrality, Katz centrality, Katz prestige and PageRank. We prove that each of these four centrality measures can be uniquely characterized with a subset of our axioms. Our system is the first one in the literature that considers all four feedback centralities.

We propose Dr.Aid, a logic-based AI framework for automated compliance checking of data governance rules over data-flow graphs. The rules are modelled using a formal language based on situation calculus and are suitable for decentralized contexts with multi-input-multi-output (MIMO) processes. Dr.Aid models data rules and flow rules and checks compliance by reasoning about the propagation, combination, modification and application of data rules over the data flow graphs. Our approach is driven and evaluated by real-world datasets using provenance graphs from data-intensive research.

CVaR-sensitive online portfolio selection (CS-OLPS) becomes increasingly important for investors because of its effectiveness to minimize conditional value at risk (CVaR) and control extreme losses. However, the non-stationary nature of financial markets makes it very difficult to address the CS-OLPS problem effectively. To address the CS-OLPS problem in non-stationary markets, we propose an effective news-driven method, named CAND, which adaptively exploits news to determine the adjustment tendency and adjustment scale for tracking the dynamic optimal portfolio with minimal CVaR in each trading round. In addition, we devise a filtering mechanism to reduce the errors caused by the noisy news for further improving CAND's effectiveness. We rigorously prove a sub-linear regret of CAND. Extensive experiments on three real-world datasets demonstrate CAND’s superiority over the state-of-the-art portfolio methods in terms of returns and risks.

Many collective decision-making settings feature a strategic tension between agents acting out of individual self-interest and promoting a common good. These include wearing face masks during a pandemic, voting, and vaccination. Networked public goods games capture this tension, with networks encoding strategic interdependence among agents. Conventional models of public goods games posit solely individual self-interest as a motivation, even though altruistic motivations have long been known to play a significant role in agents' decisions. We introduce a novel extension of public goods games to account for altruistic motivations by adding a term in the utility function that incorporates the perceived benefits an agent obtains from the welfare of others, mediated by an altruism graph. Most importantly, we view altruism not as immutable, but rather as a lever for promoting the common good. Our central algorithmic question then revolves around the computational complexity of modifying the altruism network to achieve desired public goods game investment profiles. We first show that the problem can be solved using linear programming when a principal can fractionally modify the altruism network. While the problem becomes in general intractable if the principal's actions are all-or-nothing, we exhibit several tractable special cases.

We study the allocation of indivisible goods among groups of agents using well-known fairness notions such as envy-freeness and proportionality. While these notions cannot always be satisfied, we provide several bounds on the optimal relaxations that can be guaranteed. For instance, our bounds imply that when the number of groups is constant and the $n$ agents are divided into groups arbitrarily, there exists an allocation that is envy-free up to $\Theta(\sqrt{n})$ goods, and this bound is tight. Moreover, we show that while such an allocation can be found efficiently, it is NP-hard to compute an allocation that is envy-free up to $o(\sqrt{n})$ goods even when a fully envy-free allocation exists. Our proofs make extensive use of tools from discrepancy theory.

This paper reviews current AI solutions towards road traffic congestion alleviation. Three specific AI technologies are discussed, (1) intersection management protocols for coordinating vehicles through a roads intersection in a safe and efficient manner, (2) road pricing protocol that induce optimized traffic flow, and (3) partial or full autonomous driving that can stabilize traffic flow and mitigate adverse traffic shock waves. The paper briefly presents the challenges affiliated with each of these applications along with an overview of state-of-the-art solutions. Finally, real-world implementation gaps and challenges are discussed.

Age progression and regression aim to synthesize photorealistic appearance of a given face image with aging and rejuvenation effects, respectively. Existing generative adversarial networks (GANs) based methods suffer from the following three major issues: 1) unstable training introducing strong ghost artifacts in the generated faces, 2) unpaired training leading to unexpected changes in facial attributes such as genders and races, and 3) non-bijective age mappings increasing the uncertainty in the face transformation. To overcome these issues, this paper proposes a novel framework, termed AgeFlow, to integrate the advantages of both flow-based models and GANs. The proposed AgeFlow contains three parts: an encoder that maps a given face to a latent space through an invertible neural network, a novel invertible conditional translation module (ICTM) that translates the source latent vector to target one, and a decoder that reconstructs the generated face from the target latent vector using the same encoder network; all parts are invertible achieving bijective age mappings. The novelties of ICTM are two-fold. First, we propose an attribute-aware knowledge distillation to learn the manipulation direction of age progression while keeping other unrelated attributes unchanged, alleviating unexpected changes in facial attributes. Second, we propose to use GANs in the latent space to ensure the learned latent vector indistinguishable from the real ones, which is much easier than traditional use of GANs in the image domain. Experimental results demonstrate superior performance over existing GANs-based methods on two benchmarked datasets. The source code is available at https://github.com/Hzzone/AgeFlow.

The large model size, high computational operations, and vulnerability against membership inference attack (MIA) have impeded deep learning or deep neural networks (DNNs) popularity, especially on mobile devices. To address the challenge, we envision that the weight pruning technique will help DNNs against MIA while reducing model storage and computational operation. In this work, we propose a pruning algorithm, and we show that the proposed algorithm can find a subnetwork that can prevent privacy leakage from MIA and achieves competitive accuracy with the original DNNs. We also verify our theoretical insights with experiments. Our experimental results illustrate that the attack accuracy using model compression is up to 13.6% and 10% lower than that of the baseline and Min-Max game, accordingly.

Unsupervised domain adaptation (UDA) has been received increasing attention since it does not require labels in target domain. Most existing UDA methods learn domain-invariant features by minimizing discrepancy distance computed by a certain metric between domains. However, these discrepancy-based methods cannot be robustly applied to unsupervised time series domain adaptation (UTSDA). That is because discrepancy metrics in these methods contain only low-order and local statistics, which have limited expression for time series distributions and therefore result in failure of domain matching. Actually, the real-world time series are always non-local distributions, i.e., with non-stationary and non-monotonic statistics. In this paper, we propose an Adversarial Spectral Kernel Matching (AdvSKM) method, where a hybrid spectral kernel network is specifically designed as inner kernel to reform the Maximum Mean Discrepancy (MMD) metric for UTSDA. The hybrid spectral kernel network can precisely characterize non-stationary and non-monotonic statistics in time series distributions. Embedding hybrid spectral kernel network to MMD not only guarantees precise discrepancy metric but also benefits domain matching. Besides, the differentiable architecture of the spectral kernel network enables adversarial kernel learning, which brings more discriminatory expression for discrepancy matching. The results of extensive experiments on several real-world UTSDA tasks verify the effectiveness of our proposed method.

Most existing person re-identification methods are effective in short-term scenarios because of their appearance dependencies. However, these methods may fail in long-term scenarios where people might change their clothes. To this end, we propose an adversarial feature disentanglement network (AFD-Net) which contains intra-class reconstruction and inter-class adversary to disentangle the identity-related and identity-unrelated (clothing) features. For intra-class reconstruction, the person images with the same identity are represented and disentangled into identity and clothing features by two separate encoders, and further reconstructed into original images to reduce intra-class feature variations. For inter-class adversary, the disentangled features across different identities are exchanged and recombined to generate adversarial clothes-changing images for training, which makes the identity and clothing features more independent. Especially, to supervise these new generated clothes-changing images, a re-feeding strategy is designed to re-disentangle and reconstruct these new images for image-level self-supervision in the original image space and feature-level soft-supervision in the disentangled feature space. Moreover, we collect a challenging Market-Clothes dataset and a real-world PKU-Market-Reid dataset for evaluation. The results on one large-scale short-term dataset (Market-1501) and five long-term datasets (three public and two we proposed) confirm the superiority of our method against other state-of-the-art methods.

Although deep neural networks (DNNs) have already made fairly high achievements and a very wide range of impact, their vulnerability attracts lots of interest of researchers towards related studies about artificial intelligence (AI) safety and robustness this year. A series of works reveals that the current DNNs are always misled by elaborately designed adversarial examples. And unfortunately, this peculiarity also affects real-world AI applications and places them at potential risk. we are more interested in physical attacks due to their implementability in the real world. The study of physical attacks can effectively promote the application of AI techniques, which is of great significance to the security development of AI.

Deep neural networks, particularly face recognition models, have been shown to be vulnerable to both digital and physical adversarial examples. However, existing adversarial examples against face recognition systems either lack transferability to black-box models, or fail to be implemented in practice. In this paper, we propose a unified adversarial face generation method - Adv-Makeup, which can realize imperceptible and transferable attack under the black-box setting. Adv-Makeup develops a task-driven makeup generation method with the blending module to synthesize imperceptible eye shadow over the orbital region on faces. And to achieve transferability, Adv-Makeup implements a fine-grained meta-learning based adversarial attack strategy to learn more vulnerable or sensitive features from various models. Compared to existing techniques, sufficient visualization results demonstrate that Adv-Makeup is capable to generate much more imperceptible attacks under both digital and physical scenarios. Meanwhile, extensive quantitative experiments show that Adv-Makeup can significantly improve the attack success rate under black-box setting, even attacking commercial systems.

Machine Learning (ML) increasingly informs the allocation of opportunities to individuals and communities in areas such as lending, education, employment, and beyond. Such decisions often impact their subjects' future characteristics and capabilities in an a priori unknown fashion. The decision-maker, therefore, faces exploration-exploitation dilemmas akin to those in multi-armed bandits. Following prior work, we model communities as arms. To capture the long-term effects of ML-based allocation decisions, we study a setting in which the reward from each arm evolves every time the decision-maker pulls that arm. We focus on reward functions that are initially increasing in the number of pulls but may become (and remain) decreasing after a certain point. We argue that an acceptable sequential allocation of opportunities must take an arm's potential for growth into account. We capture these considerations through the notion of policy regret, a much stronger notion than the often-studied external regret, and present an algorithm with provably sub-linear policy regret for sufficiently long time horizons. We empirically compare our algorithm with several baselines and find that it consistently outperforms them, in particular for long time horizons.

Protein secondary structure prediction (PSSP) is essential for protein function analysis. However, for low homologous proteins, the PSSP suffers from insufficient input features. In this paper, we explicitly import external self-supervised knowledge for low homologous PSSP under the guidance of residue-wise (amino acid wise) profile fusion. In practice, we firstly demonstrate the superiority of profile over Position-Specific Scoring Matrix (PSSM) for low homologous PSSP. Based on this observation, we introduce the novel self-supervised BERT features as the pseudo profile, which implicitly involves the residue distribution in all native discovered sequences as the complementary features. Furthermore, a novel residue-wise attention is specially designed to adaptively fuse different features (i.e., original low-quality profile, BERT based pseudo profile), which not only takes full advantage of each feature but also avoids noise disturbance. Besides, the feature consistency loss is proposed to accelerate the model learning from multiple semantic levels. Extensive experiments confirm that our method outperforms state-of-the-arts (i.e., 4.7% for extremely low homologous cases on BC40 dataset).

Scientists and engineers in diverse domains need to perform expensive experiments to optimize combinatorial spaces, where each candidate input is a discrete structure (e.g., sequence, tree, graph) or a hybrid structure (mixture of discrete and continuous design variables). For example, in hardware design optimization over locations of processing cores and communication links for data transfer, design evaluation involves performing a computationally-expensive simulation. These experiments are often performed in a heuristic manner by humans and without any formal reasoning. In this paper, we first describe the key challenges in solving these problems in the framework of Bayesian optimization (BO) and our progress over the last five years in addressing these challenges. We also discuss exciting sustainability applications in domains such as electronic design automation, nanoporous materials science, biological sequence design, and electric transportation systems.

Graph matching aims at establishing correspondence between node sets of given graphs while keeping the consistency between their edge sets. However, outliers in practical scenarios and equivalent learning of edge representations in deep learning methods are still challenging. To address these issues, we present an Edge Attention-adaptive Graph Matching (EAGM) network and a novel description of edge features. EAGM transforms the matching relation between two graphs into a node and edge classification problem over their assignment graph. To explore the potential of edges, EAGM learns edge attention on the assignment graph to 1) reveal the impact of each edge on graph matching, as well as 2) adjust the learning of edge representations adaptively. To alleviate issues caused by the outliers, we describe an edge by aggregating the semantic information over the space spanned by the edge. Such rich information provides clear distinctions between different edges (e.g., inlier-inlier edges vs. inlier-outlier edges), which further distinguishes outliers in the view of their associated edges. Extensive experiments demonstrate that EAGM achieves promising matching quality compared with state-of-the-arts, on cases both with and without outliers. Our source code along with the experiments is available at https://github.com/bestwei/EAGM.

As cyberattacks caused by malware have proliferated during the pandemic, building an automatic system to detect COVID-19 themed malware in social coding platforms is in urgent need. The existing methods mainly rely on file content analysis while ignoring structured information among entities in social coding platforms. Additionally, they usually require sufficient data for model training, impairing their performances over cases with limited data which is common in reality. To address these challenges, we develop Meta-AHIN, a novel model for COVID-19 themed malicious repository detection in GitHub. In Meta-AHIN, we first construct an attributed heterogeneous information network (AHIN) to model the code content and social coding properties in GitHub; and then we exploit attention-based graph convolutional neural network (AGCN) to learn repository embeddings and present a meta-learning framework for model optimization. To utilize unlabeled information in AHIN and to consider task influence of different types of repositories, we further incorporate node attribute-based self-supervised module and task-aware attention weight into AGCN and meta-learning respectively. Extensive experiments on the collected data from GitHub demonstrate that Meta-AHIN outperforms state-of-the-art methods.

A number of studies point out that current Visual Question Answering (VQA) models are severely affected by the language prior problem, which refers to blindly making predictions based on the language shortcut. Some efforts have been devoted to overcoming this issue with delicate models. However, there is no research to address it from the view of the answer feature space learning, despite the fact that existing VQA methods all cast VQA as a classification task. Inspired by this, in this work, we attempt to tackle the language prior problem from the viewpoint of the feature space learning. An adapted margin cosine loss is designed to discriminate the frequent and the sparse answer feature space under each question type properly. In this way, the limited patterns within the language modality can be largely reduced to eliminate the language priors. We apply this loss function to several baseline models and evaluate its effectiveness on two VQA-CP benchmarks. Experimental results demonstrate that our proposed adapted margin cosine loss can enhance the baseline models with an absolute performance gain of 15\% on average, strongly verifying the potential of tackling the language prior problem in VQA from the angle of the answer feature space learning.