AAAI 2023 Papers

This paper makes two key contributions. First, it argues that highly specialized rare content classifiers trained on small data typically have limited exposure to the richness and topical diversity of the negative class (dubbed anticontent) as observed in the wild. As a result, these classifiers' strong performance observed on the test set may not translate into real-world settings. In the context of COVID-19 misinformation detection, we conduct an in-the-wild audit of multiple datasets and demonstrate that models trained with several prominently cited recent datasets are vulnerable to anticontent when evaluated in the wild. Second, we present a novel active learning pipeline that requires zero manual annotation and iteratively augments the training data with challenging anticontent, robustifying these classifiers.

Spatial audio, which focuses on immersive 3D sound rendering, is widely applied in the acoustic industry. One of the key problems of current spatial audio rendering methods is the lack of personalization based on different anatomies of individuals, which is essential to produce accurate sound source positions. In this work, we address this problem from an interdisciplinary perspective. The rendering of spatial audio is strongly correlated with the 3D shape of human bodies, particularly ears. To this end, we propose to achieve personalized spatial audio by reconstructing 3D human ears with single-view images. First, to benchmark the ear reconstruction task, we introduce AudioEar3D, a high-quality 3D ear dataset consisting of 112 point cloud ear scans with RGB images. To self-supervisedly train a reconstruction model, we further collect a 2D ear dataset composed of 2,000 images, each one with manual annotation of occlusion and 55 landmarks, named AudioEar2D. To our knowledge, both datasets have the largest scale and best quality of their kinds for public use. Further, we propose AudioEarM, a reconstruction method guided by a depth estimation network that is trained on synthetic data, with two loss functions tailored for ear data. Lastly, to fill the gap between the vision and acoustics community, we develop a pipeline to integrate the reconstructed ear mesh with an off-the-shelf 3D human body and simulate a personalized Head-Related Transfer Function (HRTF), which is the core of spatial audio rendering. Code and data are publicly available in https://github.com/seanywang0408/AudioEar.

We propose a self-supervised learning approach for videos that learns representations of both the RGB frames and the accompanying audio without human supervision. In contrast to images that capture the static scene appearance, videos also contain sound and temporal scene dynamics. To leverage the temporal and aural dimension inherent to videos, our method extends temporal self-supervision to the audio-visual setting and integrates it with multi-modal contrastive objectives. As temporal self-supervision, we pose playback speed and direction recognition in both modalities and propose intra- and inter-modal temporal ordering tasks. Furthermore, we design a novel contrastive objective in which the usual pairs are supplemented with additional sample-dependent positives and negatives sampled from the evolving feature space. In our model, we apply such losses among video clips and between videos and their temporally corresponding audio clips. We verify our model design in extensive ablation experiments and evaluate the video and audio representations in transfer experiments to action recognition and retrieval on UCF101 and HMBD51, audio classification on ESC50, and robust video fingerprinting on VGG-Sound, with state-of-the-art results.

Graph Contrastive Learning (GCL) has drawn much research interest due to its strong ability to capture both graph structure and node attribute information in a self-supervised manner. Current GCL methods usually adopt Graph Neural Networks (GNNs) as the base encoder, which typically relies on the homophily assumption of networks and overlooks node similarity in the attribute space. There are many scenarios where such assumption cannot be satisfied, or node similarity plays a crucial role. In order to design a more robust mechanism, we develop a novel attribute and structure preserving graph contrastive learning framework, named ASP, which comprehensively and efficiently preserves node attributes while exploiting graph structure. Specifically, we consider three different graph views in our framework, i.e., original view, attribute view, and global structure view. Then, we perform contrastive learning across three views in a joint fashion, mining comprehensive graph information. We validate the effectiveness of the proposed framework on various real-world networks with different levels of homophily. The results demonstrate the superior performance of our model over the representative baselines.

Self-supervised anomaly detection and localization are critical to real-world scenarios in which collecting anomalous samples and pixel-wise labeling is tedious or infeasible, even worse when a wide variety of unseen anomalies could surface at test time. Our approach involves a pretext task in the context of masked image modeling, where the goal is to impose agreement between cluster assignments obtained from the representation of an image view containing saliency-aware masked patches and the uncorrupted image view. We harness the self-attention map extracted from the transformer to mask non-salient image patches without destroying the crucial structure associated with the foreground object. Subsequently, the pre-trained model is fine-tuned to detect and localize simulated anomalies generated under the guidance of the transformer's self-attention map. We conducted extensive validation and ablations on the benchmark of industrial images and achieved superior performance against competing methods. We also show the adaptability of our method to the medical images of the chest X-rays benchmark.

Monocular 3D object detection is a low-cost but challenging task, as it requires generating accurate 3D localization solely from a single image input. Recent developed depth-assisted methods show promising results by using explicit depth maps as intermediate features, which are either precomputed by monocular depth estimation networks or jointly evaluated with 3D object detection. However, inevitable errors from estimated depth priors may lead to misaligned semantic information and 3D localization, hence resulting in feature smearing and suboptimal predictions. To mitigate this issue, we propose ADD, an Attention-based Depth knowledge Distillation framework with 3D-aware positional encoding. Unlike previous knowledge distillation frameworks that adopt stereo- or LiDAR-based teachers, we build up our teacher with identical architecture as the student but with extra ground-truth depth as input. Credit to our teacher design, our framework is seamless, domain-gap free, easily implementable, and is compatible with object-wise ground-truth depth. Specifically, we leverage intermediate features and responses for knowledge distillation. Considering long-range 3D dependencies, we propose 3D-aware self-attention and target-aware cross-attention modules for student adaptation. Extensive experiments are performed to verify the effectiveness of our framework on the challenging KITTI 3D object detection benchmark. We implement our framework on three representative monocular detectors, and we achieve state-of-the-art performance with no additional inference computational cost relative to baseline models. Our code is available at https://github.com/rockywind/ADD.

The real-world Facial Expression Recognition (FER) datasets usually exhibit complex scenarios with coupled noise annotations and imbalanced classes distribution, which undoubtedly impede the development of FER methods. To address the aforementioned issues, in this paper, we propose a novel and flexible method to spot noisy labels by leveraging adversarial attack, termed as Geometry Aware Adversarial Vulnerability Estimation (GAAVE). Different from existing state-of-the-art methods of noisy label learning (NLL), our method has no reliance on additional information and is thus easy to generalize to the large-scale real-world FER datasets. Besides, the combination of Dataset Splitting module and Subset Refactoring module mitigates the impact of class imbalance, and the Self-Annotator module facilitates the sufficient use of all training data. Extensive experiments on RAF-DB, FERPlus, AffectNet, and CIFAR-10 datasets validate the effectiveness of our method. The stabilized enhancement based on different methods demonstrates the flexibility of our proposed GAAVE.

Event cameras are bio-inspired sensors that produce sparse and asynchronous event streams instead of frame-based images at a high-rate. Recent works utilizing graph convolutional networks (GCNs) have achieved remarkable performance in recognition tasks, which model event stream as spatio-temporal graph. However, the computational mechanism of graph convolution introduces redundant computation when aggregating neighbor features, which limits the low-latency nature of the events. And they perform a synchronous inference process, which can not achieve a fast response to the asynchronous event signals. This paper proposes a local-shift graph convolutional network (LSNet), which utilizes a novel local-shift operation equipped with a local spatio-temporal attention component to achieve efficient and adaptive aggregation of neighbor features. To improve the efficiency of pooling operation in feature extraction, we design a node-importance based parallel pooling method (NIPooling) for sparse and low-latency event data. Based on the calculated importance of each node, NIPooling can efficiently obtain uniform sampling results in parallel, which retains the diversity of event streams. Furthermore, for achieving a fast response to asynchronous event signals, an asynchronous event processing procedure is proposed to restrict the network nodes which need to recompute activations only to those affected by the new arrival event. Experimental results show that the computational cost can be reduced by nearly 9 times through using local-shift operation and the proposed asynchronous procedure can further improve the inference efficiency, while achieving state-of-the-art performance on gesture recognition and object recognition.

While neuromorphic computing architectures based on Spiking Neural Networks (SNNs) are increasingly gaining interest as a pathway toward bio-plausible machine learning, attention is still focused on computational units like the neuron and synapse. Shifting from this neuro-synaptic perspective, this paper attempts to explore the self-repair role of glial cells, in particular, astrocytes. The work investigates stronger correlations with astrocyte computational neuroscience models to develop macro-models with a higher degree of bio-fidelity that accurately captures the dynamic behavior of the self-repair process. Hardware-software co-design analysis reveals that bio-morphic astrocytic regulation has the potential to self-repair hardware realistic faults in neuromorphic hardware systems with significantly better accuracy and repair convergence for unsupervised learning tasks on the MNIST and F-MNIST datasets. Our implementation source code and trained models are available at https://github.com/NeuroCompLab-psu/Astromorphic_Self_Repair.

My dissertation research focuses on sequential decision-making (SDM) in complex environments, and how agents can perform well even when novelty is introduced to those environments. The problem of how agents can respond intelligently to novelty has been a long-standing challenge in AI, and poses unique problems across approaches to SDM. This question has been studied in various formulations, including open-world learning and reasoning, transfer learning, concept drift, and statistical relational learning. Classical and modern approaches in agent design offer tradeoffs in human effort for feature encoding, ease of deployment in new domains, and the development of both provably and empirically reliable policies. I propose a formalism for studying open-world novelty in SDM processes with feature-rich observations. I study the conditions under which causal-relational queries can be estimated from non-novel observations, and empirically examine the effects of open-world novelty on agent behavior.

Early diagnosis of osteonecrosis of the femoral head (ONFH) can inhibit the progression and improve femoral head preservation. The radiograph difference between early ONFH and healthy ones is not apparent to the naked eye. It is also hard to produce a large dataset to train the classification model. In this paper, we propose Asymmetric-Sensitive Transformer (AsT) to capture the uneven development of the bilateral femoral head to enable robust ONFH detection. Our ONFH detection is realized using the self-attention mechanism to femoral head regions while conferring sensitivity to the uneven development by the attention-shared transformer. The real-world experiment studies show that AsT achieves the best performance of AUC 0.9313 in the early diagnosis of ONFH and can find out misdiagnosis cases firmly.

Argument(ation) mining (AM) is an area of research in Artificial Intelligence (AI) that aims to identify, analyse and automatically generate arguments in natural language. In a pipeline, the identification and analysis of the arguments and their components (i.e. premises and claims) in texts and the prediction of their relations (i.e. attack and support) are then handled by argument-based reasoning frameworks so that, for example, fallacies and inconsistencies can be automatically identified. Recently, the field of argument mining has tackled new challenges, namely the evaluation of argument quality (e.g. strength, persuasiveness), natural language argument summarisation and retrieval, and natural language argument generation. In this paper, I discuss my main contributions in this area as well as some lines of future research. This paper is part of the AAAI-23 New Faculty Highlights.

Recently, there has been a surge of Transformer-based solutions for the long-term time series forecasting (LTSF) task. Despite the growing performance over the past few years, we question the validity of this line of research in this work. Specifically, Transformers is arguably the most successful solution to extract the semantic correlations among the elements in a long sequence. However, in time series modeling, we are to extract the temporal relations in an ordered set of continuous points. While employing positional encoding and using tokens to embed sub-series in Transformers facilitate preserving some ordering information, the nature of the permutation-invariant self-attention mechanism inevitably results in temporal information loss. To validate our claim, we introduce a set of embarrassingly simple one-layer linear models named LTSF-Linear for comparison. Experimental results on nine real-life datasets show that LTSF-Linear surprisingly outperforms existing sophisticated Transformer-based LTSF models in all cases, and often by a large margin. Moreover, we conduct comprehensive empirical studies to explore the impacts of various design elements of LTSF models on their temporal relation extraction capability. We hope this surprising finding opens up new research directions for the LTSF task. We also advocate revisiting the validity of Transformer-based solutions for other time series analysis tasks (e.g., anomaly detection) in the future.

We study a general allocation setting where agent valuations are concave additive. In this model, a collection of items must be uniquely distributed among a set of agents, where each agent-item pair has a specified utility. The objective is to maximize the sum of agent valuations, each of which is an arbitrary non-decreasing concave function of the agent's total additive utility. This setting was studied by Devanur and Jain (STOC 2012) in the online setting for divisible items. In this paper, we obtain both multiplicative and additive approximations in the offline setting for indivisible items. Our approximations depend on novel parameters that measure the local multiplicative/additive curvatures of each agent valuation, which we show correspond directly to the integrality gap of the natural assignment convex program of the problem. Furthermore, we extend our additive guarantees to obtain constant multiplicative approximations for Asymmetric Nash Welfare Maximization when agents have smooth valuations. This algorithm also yields an interesting tatonnement-style interpretation, where agents adjust uniform prices and items are assigned according to maximum weighted bang-per-buck ratios.

Conformal prediction (CP) is a wrapper around traditional machine learning models, giving coverage guarantees under the sole assumption of exchangeability; in classification problems, a CP guarantees that the error rate is at most a chosen significance level, irrespective of whether the underlying model is misspecified. However, the prohibitive computational costs of full CP led researchers to design scalable alternatives, which alas do not attain the same guarantees or statistical power of full CP. In this paper, we use influence functions to efficiently approximate full CP. We prove that our method is a consistent approximation of full CP, and empirically show that the approximation error becomes smaller as the training set increases; e.g., for 1,000 training points the two methods output p-values that are <0.001 apart: a negligible error for any practical application. Our methods enable scaling full CP to large real-world datasets. We compare our full CP approximation (ACP) to mainstream CP alternatives, and observe that our method is computationally competitive whilst enjoying the statistical predictive power of full CP.

We consider a voting scenario in which the resource to be voted upon may consist of both indivisible and divisible goods. This generalizes both the well-studied model of multiwinner voting and the recently introduced model of cake sharing. Under approval votes, we propose two variants of the extended justified representation (EJR) notion from multiwinner voting, a stronger one called EJR for mixed goods (EJR-M) and a weaker one called EJR up to 1 (EJR-1). We extend three multiwinner voting rules to our setting—GreedyEJR, the method of equal shares (MES), and proportional approval voting (PAV)—and show that while all three generalizations satisfy EJR-1, only the first one provides EJR-M. In addition, we derive tight bounds on the proportionality degree implied by EJR-M and EJR-1, and investigate the proportionality degree of our proposed rules.

Predicting user engagement -- whether a user will engage in a given information cascade -- is an important problem in the context of social media, as it is useful to online marketing and misinformation mitigation just to name a couple major applications. Based on split population multi-variate survival processes, we develop a discriminative approach that, unlike prior works, leads to a single model for predicting whether individual users of an information network will engage a given cascade for arbitrary forecast horizons and observation periods. Being probabilistic in nature, this model retains the interpretability of its generative counterpart and renders count prediction intervals in a disciplined manner. Our results indicate that our model is highly competitive, if not superior, to current approaches, when compared over varying observed cascade histories and forecast horizons.

Feature selection (FS) is a crucial procedure in machine learning pipelines for its significant benefits in removing data redundancy and mitigating model overfitting. Since concept drift is a widespread phenomenon in streaming data and could severely affect model performance, effective FS on concept drifting data streams is imminent. However, existing state-of-the-art FS algorithms fail to adjust their selection strategy adaptively when the effective feature subset changes, making them unsuitable for drifting streams. In this paper, we propose a dynamic FS method that selects effective features on concept drifting data streams via deep reinforcement learning. Specifically, we present two novel designs: (i) a skip-mode reinforcement learning environment that shrinks action space size for high-dimensional FS tasks; (ii) a curiosity mechanism that generates intrinsic rewards to address the long-horizon exploration problem. The experiment results show that our proposed method outperforms other FS methods and can dynamically adapt to concept drifts.

Skeleton-based action recognition attracts practitioners and researchers due to the lightweight, compact nature of datasets. Compared with RGB-video-based action recognition, skeleton-based action recognition is a safer way to protect the privacy of subjects while having competitive recognition performance. However, due to improvements in skeleton recognition algorithms as well as motion and depth sensors, more details of motion characteristics can be preserved in the skeleton dataset, leading to potential privacy leakage. We first train classifiers to categorize private information from skeleton trajectories to investigate the potential privacy leakage from skeleton datasets. Our preliminary experiments show that the gender classifier achieves 87% accuracy on average, and the re-identification classifier achieves 80% accuracy on average with three baseline models: Shift-GCN, MS-G3D, and 2s-AGCN. We propose an anonymization framework based on adversarial learning to protect potential privacy leakage from the skeleton dataset. Experimental results show that an anonymized dataset can reduce the risk of privacy leakage while having marginal effects on action recognition performance even with simple anonymizer architectures. The code used in our experiments is available at https://github.com/ml-postech/Skeleton-anonymization/