IJCAI 2022 论文列表

Graph neural networks (GNNs) have been widely used in modeling graph structured data, owing to its impressive performance in a wide range of practical applications. Recently, knowledge distillation (KD) for GNNs has enabled remarkable progress in graph model compression and knowledge transfer. However, most of the existing KD methods require a large volume of real data, which are not readily available in practice, and may preclude their applicability in scenarios where the teacher model is trained on rare or hard to acquire datasets. To address this problem, we propose the first end-to-end framework for data-free adversarial knowledge distillation on graph structured data (DFAD-GNN). To be specific, our DFAD-GNN employs a generative adversarial network, which mainly consists of three components: a pre-trained teacher model and a student model are regarded as two discriminators, and a generator is utilized for deriving training graphs to distill knowledge from the teacher model into the student model. Extensive experiments on various benchmark models and six representative datasets demonstrate that our DFAD-GNN significantly surpasses state-of-the-art data-free baselines in the graph classification task.

Recent studies have proven that deep neural networks are vulnerable to backdoor attacks. Specifically, by mixing a small number of poisoned samples into the training set, the behavior of the trained model can be maliciously controlled. Existing attack methods construct such adversaries by randomly selecting some clean data from the benign set and then embedding a trigger into them. However, this selection strategy ignores the fact that each poisoned sample contributes inequally to the backdoor injection, which reduces the efficiency of poisoning. In this paper, we formulate improving the poisoned data efficiency by the selection as an optimization problem and propose a Filtering-and-Updating Strategy (FUS) to solve it. The experimental results on CIFAR-10 and ImageNet-10 indicate that the proposed method is effective: the same attack success rate can be achieved with only 47% to 75% of the poisoned sample volume compared to the random selection strategy. More importantly, the adversaries selected according to one setting can generalize well to other settings, exhibiting strong transferability. The prototype code of our method is now available at https://github.com/xpf/Data-Efficient-Backdoor-Attacks.

Recently proposed pre-trained language models can be easily fine-tuned to a wide range of downstream tasks. However, fine-tuning requires a large training set. This PhD project introduces novel natural language processing (NLP) use cases in the healthcare domain where obtaining a large training dataset is difficult and expensive. To this end, we propose data-efficient algorithms to fine-tune NLP models in low-resource settings and validate their effectiveness. We expect the outcomes of this PhD project could contribute to the NLP research and low-resource application domains.

Data valuation in machine learning (ML) is an emerging research area that studies the worth of data in ML. Data valuation is used in collaborative ML to determine a fair compensation for every data owner and in interpretable ML to identify the most responsible, noisy, or misleading training examples. This paper presents a comprehensive technical survey that provides a new formal study of data valuation in ML through its “ingredients” and the corresponding properties, grounds the discussion of common desiderata satisfied by existing data valuation strategies on our proposed ingredients, and identifies open research challenges for designing new ingredients, data valuation strategies, and cost reduction techniques.

Improving generalization in text-based games serves as a useful stepping-stone towards reinforcement learning (RL) agents with generic linguistic ability. Data augmentation for generalization in RL has shown to be very successful in classic control and visual tasks, but there is no prior work for text-based games. We propose Transition-Matching Permutation, a novel data augmentation technique for text-based games, where we identify phrase permutations that match as many transitions in the trajectory data. We show that applying this technique results in state-of-the-art performance in the Cooking Game benchmark suite for text-based games.

The recently developed Particle-based Variational Inference (ParVI) methods drive the empirical distribution of a set of fixed-weight particles towards a given target distribution by iteratively updating particles' positions. However, the fixed weight restriction greatly confines the empirical distribution's approximation ability, especially when the particle number is limited. In this paper, we propose to dynamically adjust particles' weights according to a Fisher-Rao reaction flow. We develop a general Dynamic-weight Particle-based Variational Inference (DPVI) framework according to a novel continuous composite flow, which evolves the positions and weights of particles simultaneously. We show that the mean-field limit of our composite flow is actually a Wasserstein-Fisher-Rao gradient flow of the associated dissimilarity functional. By using different finite-particle approximations in our general framework, we derive several efficient DPVI algorithms. The empirical results demonstrate the superiority of our derived DPVI algorithms over their fixed-weight counterparts.

The problem of exact weighted sampling of solutions of Boolean formulas has applications in Bayesian inference, testing, and verification. The state-of-the-art approach to sampling involves carefully decomposing the input formula and compiling a data structure called d-DNNF in the process. Recent work in the closely connected field of model counting, however, has shown that smartly composing different subformulas using dynamic programming and Algebraic Decision Diagrams (ADDs) can outperform d-DNNF-style approaches on many benchmarks. In this work, we present a modular algorithm called DPSampler that extends such dynamic-programming techniques to the problem of exact weighted sampling. DPSampler operates in three phases. First, an execution plan in the form of a project-join tree is computed using tree decompositions. Second, the plan is used to compile the input formula into a succinct tree-of-ADDs representation. Third, this tree is traversed to generate a random sample. This decoupling of planning, compilation and sampling phases enables usage of specialized libraries for each purpose in a black-box fashion. Further, our novel ADD-sampling algorithm avoids the need for expensive dynamic memory allocation required in previous work. Extensive experiments over diverse sets of benchmarks show DPSampler is more scalable and versatile than existing approaches.

Despite their outstanding performance in a broad spectrum of real-world tasks, deep artificial neural networks are sensitive to input noises, particularly adversarial perturbations. On the contrary, human and animal brains are much less vulnerable. In contrast to the one-shot inference performed by most deep neural networks, the brain often solves decision-making with an evidence accumulation mechanism that may trade time for accuracy when facing noisy inputs. The mechanism is well described by the Drift-Diffusion Model (DDM). In the DDM, decision-making is modeled as a process in which noisy evidence is accumulated toward a threshold. Drawing inspiration from the DDM, we propose the Dropout-based Drift-Diffusion Model (DDDM) that combines test-phase dropout and the DDM for improving the robustness for arbitrary neural networks. The dropouts create temporally uncorrelated noises in the network that counter perturbations, while the evidence accumulation mechanism guarantees a reasonable decision accuracy. Neural networks enhanced with the DDDM tested in image, speech, and text classification tasks all significantly outperform their native counterparts, demonstrating the DDDM as a task-agnostic defense against adversarial attacks.

Rain streaks and background components in a rainy input are highly correlated, making the deraining task a composition of the rain streak removal and background restoration. However, the correlation of these two components is barely considered, leading to unsatisfied deraining results. To this end, we propose a dynamic associated network (DANet) to achieve the association learning between rain streak removal and background recovery. There are two key aspects to fulfill the association learning: 1) DANet unveils the latent association knowledge between rain streak prediction and background texture recovery, and leverages it as an extra prior via an associated learning module (ALM) to promote the texture recovery. 2) DANet introduces the parametric association constraint for enhancing the compatibility of deraining model with background reconstruction, enabling it to be automatically learned from the training data. Moreover, we observe that the sampled rainy image enjoys the similar distribution to the original one. We thus propose to learn the rain distribution at the sampling space, and exploit super-resolution to reconstruct high-frequency background details for computation and memory reduction. Our proposed DANet achieves the approximate deraining performance to the state-of-the-art MPRNet but only requires 52.6\% and 23\% inference time and computational cost, respectively.

The non-uniformly distributed nature of the 3D Dynamic Point Cloud (DPC) brings significant challenges to its high-efficient inter-frame compression. This paper proposes a novel 3D sparse convolution-based Deep Dynamic Point Cloud Compression (D-DPCC) network to compensate and compress the DPC geometry with 3D motion estimation and motion compensation in the feature space. In the proposed D-DPCC network, we design a Multi-scale Motion Fusion (MMF) module to accurately estimate the 3D optical flow between the feature representations of adjacent point cloud frames. Specifically, we utilize a 3D sparse convolution-based encoder to obtain the latent representation for motion estimation in the feature space and introduce the proposed MMF module for fused 3D motion embedding. Besides, for motion compensation, we propose a 3D Adaptively Weighted Interpolation (3DAWI) algorithm with a penalty coefficient to adaptively decrease the impact of distant neighbours. We compress the motion embedding and the residual with a lossy autoencoder-based network. To our knowledge, this paper is the first work proposing an end-to-end deep dynamic point cloud compression framework. The experimental result shows that the proposed D-DPCC framework achieves an average 76% BD-Rate (Bjontegaard Delta Rate) gains against state-of-the-art Video-based Point Cloud Compression (V-PCC) v13 in inter mode.

Despite the success of text-to-text pre-trained models in various natural language generation (NLG) tasks, the generation performance is largely restricted by the number of labeled data in downstream tasks, particularly in data-to-text generation tasks. Existing works mostly utilize abundant unlabeled structured data to conduct unsupervised pre-training for task adaption, which fail to model the complex relationship between source structured data and target texts. Thus, we introduce self-training as a better few-shot learner than task-adaptive pre-training, which explicitly captures this relationship via pseudo-labeled data generated by the pre-trained model. To alleviate the side-effect of low-quality pseudo-labeled data during self-training, we propose a novel method called Curriculum-Based Self-Training (CBST) to effectively leverage unlabeled data in a rearranged order determined by the difficulty of text generation. Experimental results show that our method can outperform fine-tuning and task-adaptive pre-training methods, and achieve state-of-the-art performance in the few-shot setting of data-to-text generation.

We address the problem of predicting when a disease will develop, i.e., medical event time (MET), from a patient's electronic health record (EHR). The MET of non-communicable diseases like diabetes is highly correlated to cumulative health conditions, more specifically, how much time the patient spent with specific health conditions in the past. The common time-series representation is indirect in extracting such information from EHR because it focuses on detailed dependencies between values in successive observations, not cumulative information. We propose a novel data representation for EHR called cumulative stay-time representation (CTR), which directly models such cumulative health conditions. We derive a trainable construction of CTR based on neural networks that has the flexibility to fit the target data and scalability to handle high-dimensional EHR. Numerical experiments using synthetic and real-world datasets demonstrate that CTR alone achieves a high prediction performance, and it enhances the performance of existing models when combined with them.

Crowd counting methods typically predict a density map as an intermediate representation of counting, and achieve good performance. However, due to the perspective phenomenon, there is a scale variation in real scenes, which causes the density map-based methods suffer from a severe scene generalization problem because only a limited number of scales are fitted in density map prediction and generation. To address this issue, we propose a novel vision transformer network, i.e., CrowdFormer, and a density kernels fusion framework for more accurate density map estimation and generation, respectively. Thereafter, we incorporate these two innovations into an adaptive learning system, which can take both the annotation dot map and original image as input, and jointly learns the density map estimator and generator within an end-to-end framework. The experimental results demonstrate that the proposed model achieves the state-of-the-art in the terms of MAE and MSE (e.g., it achieved a MAE of 67.1 and MSE of 301.6 on NWPU-Crowd dataset.), and confirm the effectiveness of the proposed two designs. The code is https://github.com/special-yang/Top_Down-CrowdCounting.

In this paper, we study an explainable COVID-19 misinformation detection problem where the goal is to accurately identify COVID-19 misleading posts on social media and explain the posts with natural language explanations (NLEs). Our problem is motivated by the limitations of current explainable misinformation detection approaches that cannot provide NLEs for COVID-19 posts due to the lack of sufficient professional COVID-19 knowledge for supervision. To address such a limitation, we develop CEA-COVID, a crowd-expert-AI framework that jointly exploits the common logical reasoning ability of online crowd workers and the professional knowledge of COVID-19 experts to effectively generate NLEs for detecting and explaining COVID-19 misinformation. We evaluate CEA-COVID using two public COVID-19 misinformation datasets on social media. Results demonstrate that CEA-COVID outperforms existing explainable misinformation detection models in terms of both explainability and detection accuracy.

Since single-modal controllable manipulation typically requires supervision of information from other modalities or cooperation with complex software and experts, this paper addresses the problem of cross-modal adaptive manipulation (CAM). The novel task performs cross-modal semantic alignment from mutual supervision and implements bidirectional exchange of attributes, relations, or objects in parallel, benefiting both modalities while significantly reducing manual effort. We introduce a robust solution for CAM, which includes two essential modules, namely Heterogeneous Representation Learning (HRL) and Cross-modal Relation Reasoning (CRR). The former is designed to perform representation learning for cross-modal semantic alignment on heterogeneous graph nodes. The latter is adopted to identify and exchange the focused attributes, relations, or objects in both modalities. Our method produces pleasing cross-modal outputs on CUB and Visual Genome.

Every year there are 1.9 million deaths world-wide attributed to occupational health and safety risk factors. To address poor working conditions and fulfill UN's SDG 8, "protect labour rights and promote safe working environments for all workers", governmental agencies conduct labour inspections, using checklists to survey individual organisations for working environment violations. Recent research highlights the benefits of using machine learning for creating checklists. However, the current methods only create static checklists and do not adapt them to new information that surfaces during use. In contrast, we propose a new method called Context-aware Bayesian Case-Based Reasoning (CBCBR) that creates dynamic checklists. These checklists are continuously adapted as the inspections progress, based on how they are answered. Our evaluations show that CBCBR's dynamic checklists outperform static checklists created via the current state-of-the-art methods, increasing the expected number of working environment violations found in the labour inspections.

The current generation of symbolic reasoning techniques excel at the qualitative tasks (i.e., when the answer is Yes or No); such techniques sufficed for traditional systems whose design sought to achieve deterministic behavior. In contrast, modern computing systems crucially rely on the statistical methods to account for the uncertainty in the environment, and to reason about behavior of these systems, there is need to look beyond qualitative symbolic reasoning techniques. We will discuss our work focused on the development of the next generation of automated reasoning techniques that can perform higher-order tasks such as quantitative measurement, sampling of representative behavior, and automated synthesis of systems. From a core technical perspective, our work builds on the SAT revolution, which refers to algorithmic advances in combinatorial solving techniques for the fundamental problem of satisfiability (SAT), i.e., whether it is possible to satisfy a given set of constraints. The SAT revolution offers the opportunity to develop scalable techniques for problems that lie beyond SAT from complexity perspective and, therefore, stand to benefit from the availability of powerful SAT engines. Our work seeks to enable a Beyond SAT revolution via design of scalable techniques for three fundamental problems that lie beyond SAT: constrained counting, constrained sampling, and automated synthesis.

Bias in the training data can jeopardize fairness and explainability of deep neural network prediction on test data. We propose a novel bias-tailored data augmentation approach, Counterfactual Interpolation Augmentation (CIA), attempting to debias the training data by d-separating the spurious correlation between the target variable and the sensitive attribute. CIA generates counterfactual interpolations along a path simulating the distribution transitions between the input and its counterfactual example. CIA as a pre-processing approach enjoys two advantages: First, it couples with either plain training or debiasing training to markedly increase fairness over the sensitive attribute. Second, it enhances the explainability of deep neural networks by generating attribution maps via integrating counterfactual gradients. We demonstrate the superior performance of the CIA-trained deep neural network models using qualitative and quantitative experimental results. Our code is available at: https://github.com/qiangyao1988/CIA

Recently, many studies have tried to create generation models to assist counter speakers by providing counterspeech suggestions for combating the explosive proliferation of online hate. However, since these suggestions are from a vanilla generation model, they might not include the appropriate properties required to counter a particular hate speech instance. In this paper, we propose CounterGeDi - an ensemble of generative discriminators (GeDi) to guide the generation of a DialoGPT model toward more polite, detoxified, and emotionally laden counterspeech. We generate counterspeech using three datasets and observe significant improvement across different attribute scores. The politeness and detoxification scores increased by around 15% and 6% respectively, while the emotion in the counterspeech increased by at least 10% across all the datasets. We also experiment with triple-attribute control and observe significant improvement over single attribute results when combining complementing attributes, e.g., politeness, joyfulness and detoxification. In all these experiments, the relevancy of the generated text does not deteriorate due to the application of these controls.