Liang Chen

EAAI Journal 2026 Journal Article

A novel transfer learning method for bearing fault diagnosis based on squeeze-excitation dilated SincNet combined with physics-informed subdomain adaptation

• Jingshu Zhong
• Liang Chen
• Siqi Qiu
• Chenhan Wang
• Yu Zheng

Transfer learning methods are widely applied to rolling bearing fault diagnosis under varying working conditions, but the accuracy is influenced by transfer strategies. To achieve more effective domain alignment and eliminate irrelevant information, a novel transfer learning method based on Squeeze-and-Excitation Dilated SincNet with physics-informed subdomain adaptation is proposed. First, Squeeze-and-Excitation mechanisms and dilated convolutions are incorporated into the SincNet framework to enable adaptive sub-signal extraction and receptive field expansion. Second, based on bearing fault mechanisms, physics-informed transfer metrics characterizing impulsiveness, transient impacts, and fault frequency band correlations are established. Subsequently, Multi-dimensional alignment is achieved between source and target domains through integrated domain adaptation, subdomain adaptation, and adversarial learning modules. Validation on the Paderborn University rolling bearing dataset demonstrates that the proposed method achieves optimal performance across multiple transfer tasks.

EAAI Journal 2026 Journal Article

Deep reinforcement learning-based dynamic integrated scheduling of automated guided vehicles and yard cranes for container terminal loading operations

• Yuxuan Zhang
• Liang Chen
• Moshi Zhou
• Xiangyu Bao
• Funing Jia
• Changhui Liu
• Lei Zhang
• Yu Zheng

Integrated scheduling of automated guided vehicles (AGVs) and yard cranes (YCs) is crucial for enhancing loading efficiency in container terminals. However, most existing integrated scheduling models are deterministic and static, which limits their effectiveness in uncertain environments. This paper attempts to find reactive scheduling policies that respond to actual observed information rather than relying on determined handling and transport times. We model the loading operation with uncertain handling and transport times as a semi-open queuing network (SOQN) known as LO-SOQN. LO-SOQN responds instantly to AGV and YC scheduling decisions, which provides quantitative metrics for optimizing dynamic scheduling policies. We formulate the problem as a Markov decision process (MDP) and propose a deep reinforcement learning (DRL) approach to find near-optimal policies. The proposed DRL approach ensures policy generalization by learning uniform state representation, which allows it to be applied to flexible equipment configurations. A series of simulation experiments evaluates the performance of our approach under both fixed and flexible equipment configurations. Compared to the inventory-based and robust fluid policies, the proposed approach reduces the long-term average turnaround time by 10. 28 %–18. 94 %. Furthermore, the generalization curve indicates the feasibility and effectiveness of training a policy network that generalizes to various equipment configurations.

AAAI Conference 2026 Conference Paper

GT-SNT: A Linear-Time Transformer for Large-Scale Graphs via Spiking Node Tokenization

• Huizhe Zhang
• Jintang Li
• Yuchang Zhu
• Huazhen Zhong
• Liang Chen

Graph Transformers (GTs), which integrate message passing and self-attention mechanisms simultaneously, have achieved promising empirical results in graph prediction tasks. However, the design of scalable and topology-aware node tokenization has lagged behind other modalities. This gap becomes critical as the quadratic complexity of full attention renders them impractical on large-scale graphs. Recently, Spiking Neural Networks (SNNs), as brain-inspired models, provided an energy-saving scheme to convert input intensity into discrete spike-based representations through event-driven spiking neurons. Inspired by these characteristics, we propose a linear-time Graph Transformer with Spiking Node Tokenization (GT-SNT) for node classification. By integrating multi-step feature propagation with SNNs, spiking node tokenization generates compact, locality-aware spike count embeddings as node tokens to avoid predefined codebooks and their utilization issues. The codebook guided self-attention leverages these tokens to perform node-to-token attention for linear-time global context aggregation. In experiments, we compare GT-SNT with other state-of-the-art baselines on node classification datasets ranging from small to large. Experimental results show that GT-SNT achieves comparable performances on most datasets and reaches up to 130× faster inference speed compared to other GTs.

EAAI Journal 2026 Journal Article

Multi-objective optimization of sealing and friction performance for the cylinder liner-piston ring friction pair in highly strengthened diesel engines based on machine learning

• Yuchen Xu
• Guisheng Chen
• Liang Chen
• Size Zhang
• Yuhua Bi
• Renxin Xiao

With increasing engine boosting, cylinder liners experience higher thermal loads and peak firing pressures. This induces thermo-mechanically coupled out-of-round deformation and weakens conformity at the cylinder liner-piston ring (CLPR) interface. The resulting increase in blow-by and friction losses reduces engine mechanical efficiency, while upward oil transport into the combustion chamber can aggravate particulate emissions. This study aims to systematically optimize cylinder-liner pre-compensation geometric parameters to improve the sealing and friction performance of the CLPR system, using lubricating oil consumption (LOC), blow-by gas flow (BGF), and friction mean effective pressure (FMEP) as evaluation metrics. A multi-physics-driven and surrogate-assisted optimization framework is developed. A fully parametric pre-compensation design space is introduced, and a dataset is generated via Latin hypercube sampling (LHS) to train a Kriging surrogate model. The third-generation non-dominated sorting genetic algorithm (NSGA-III) is then applied to obtain the Pareto set, and the entropy-weighted Technique for Order Preference by Similarity to an Ideal Solution (E-TOPSIS) is used to select a representative compromise solution, which is further verified across multiple engine speeds. The results indicate that the ovality parameter is the dominant factor influencing LOC, BGF, and FMEP. Under rated conditions, the optimal pre-compensation design reduces LOC by 2. 90%, BGF by 1. 02%, and FMEP by 41. 32% relative to the baseline, and the improvements are retained over a range of engine speeds. The proposed approach provides a manufacturable and practically implementable route to mitigate deformation-induced sealing leakage and friction losses, thereby supporting mechanical-efficiency improvement and emission mitigation in highly boosted engines.

EAAI Journal 2026 Journal Article

Real-time quantitative analysis of wildfire fireline merging behavior based on segmentation-skeletonization algorithm

• Jianwei Li
• Liang Chen
• Xinxin Zhao

Multiple interacting wildfires often produce fireline merging, which can abruptly intensify fire potential and trigger high-risk phenomena such as deflagration and fire jumping. Real-time, quantitative identification of fireline merging behavior is therefore critical for scientific decision-making and efficient emergency response. In this paper, we propose an artificial intelligence framework that integrates deep-learning-based image segmentation with skeletonization analysis to quantify fireline merging behavior and mitigate the problems of lagged capture and low recognition accuracy in current fire monitoring. The framework adopts the You Only Look Once version 8 segmentation (YOLOv8-seg) model integrated with Bottleneck Transformer Block (BoTBlock), which combines convolutional feature extraction with multi-head self-attention (MHSA) to enhance global semantic perception and improve fireline segmentation accuracy and boundary consistency. Building on this, topological skeleton extraction and Basis spline (B-spline) curve smoothing are used to obtain the main fireline path, while curvature extremum points are employed to automatically locate the merging position and characterize the merging trend. Experiments show that the system achieves an inference speed of 30 frames/second and a segmentation mean Average Precision at Intersection over Union(IoU) 0. 5–0. 95 (mAP50–95) of 83. 3 %, effectively balancing real-time performance and accuracy. Comparisons with typical field measurements and published studies indicate that the proposed method can reliably recover the trajectory and key behavioral characteristics of fireline merging in dynamic fire scenes.

AAAI Conference 2026 Conference Paper

Spiking Heterogeneous Graph Attention Networks

• Buqing Cao
• Qian Peng
• Xiang Xie
• Liang Chen
• Min Shi
• Jianxun Liu

Real-world graphs or networks are usually heterogeneous, involving multiple types of nodes and relationships. Heterogeneous graph neural networks (HGNNs) can effectively handle these diverse nodes and edges, capturing heterogeneous information within the graph, thus exhibiting outstanding performance. However, most methods of HGNNs usually involve complex structural designs, leading to problems such as high memory usage, long inference time, and extensive consumption of computing resources. These limitations pose certain challenges for the practical application of HGNNs, especially for resource-constrained devices. To mitigate this issue, we propose the Spiking Heterogeneous Graph Attention Networks (SpikingHAN), which incorporates the brain-inspired and energy-saving properties of Spiking Neural Networks (SNNs) into heterogeneous graph learning to reduce the computing cost without compromising the performance. Specifically, SpikingHAN aggregates metapath-based neighbor information using a single-layer graph convolution with shared parameters. It then employs a semantic-level attention mechanism to capture the importance of different meta-paths and performs semantic aggregation. Finally, it encodes the heterogeneous information into a spike sequence through SNNs, simulating bioinformatic processing to derive a binarized 1-bit representation of the heterogeneous graph. Comprehensive experimental results from three real-world heterogeneous graph datasets show that SpikingHAN delivers competitive node classification performance. It achieves this with fewer parameters, quicker inference, reduced memory usage, and lower energy consumption.

TCS Journal 2025 Journal Article

Cyclic diagnosability of folded hypercubes under the PMC model and MM* model

• Linxiao Wang
• Liang Chen
• Kaineng Guan
• Yanze Huang
• Limei Lin

Cyclic diagnosability ensures reliable fault detection while maintaining loop-based connectivity, which is vital for fault-tolerant systems like data centers. Unlike traditional methods, it guarantees fault identification even if the surviving network splits into at least two cyclic components, improving resilience in connectivity-critical scenarios. With the growing need for reliable interconnection networks in large-scale multiprocessor systems, this paper explores the cyclic diagnosability of n -dimensional folded hypercubes F Q n under the PMC and MM* models. Leveraging the structural advantages of F Q n, such as its complementary edges and enhanced connectivity, we develop a systematic method to determine its cyclic diagnosability. Our analysis shows that for n ≥ 10, the cyclic diagnosability c t ( F Q n ) reaches 5 n − 5, a significant improvement over the 5 n − 10 limit of classical hypercubes. This result highlights the superior fault tolerance of folded hypercubes in maintaining loop-based connectivity during fault diagnosis.

EAAI Journal 2025 Journal Article

Remaining useful life prediction of machinery using federated public feature representation in edge-cloud collaboration architecture

• Liang Chen
• Hongli Gao
• Liang Guo
• Junhua Liang
• Lin Peng

Significant progress has been made in the prediction methods of the remaining useful life (RUL) of machinery. Nevertheless, two major challenges still exist for the large-scale practical application of these methods. Firstly, most prediction methods necessitate operation on high compute capability devices, which inevitably leads to a substantial increase in costs. Secondly, there are issues related to data silos and privacy protection. To address these challenges, we propose a new RUL prediction method by using federated public feature representation in edge-cloud collaboration (ECC) architecture. The training process of the proposed method contains two stages. In the first stage, the proposed Federated Feature Separation Autoencoder (FedFSAE) extracts shallow public feature representations from data located at different clients, with similar distributions. FedFSAE only requires training a shallow autoencoder, it does not require high compute capability on the clients. Furthermore, by aggregating only the public encoders, it achieves low communication costs and enhances privacy protection. In the second stage, clients upload the public features and corresponding labels to a trusted server with high compute capability, then, a global RUL prediction model, the proposed long-term fully convolutional network (LTFCN), is trained. This fully utilizes the high compute capability of the server to obtain a high-performance prediction model. Finally, the comparative analysis was conducted on milling cutter and bearing datasets to validate the effectiveness and superiority of the proposed method.

IJCAI Conference 2025 Conference Paper

RepObE: Representation Learning-Enhanced Obfuscation Encryption Modular Semantic Task Framework

• Limei Lin
• Jinpeng Xu
• Xiaoding Wang
• Liang Chen
• Sun-Yuan Hsieh
• Jie Wu

Model inversion and adversarial attacks in semantic communication pose risks, such as content leaks, alterations, and prediction inaccuracies, which threaten security and reliability. This paper introduces, from an attacker's viewpoint, a novel framework called RepObE (Representation Learning-Enhanced Obfuscation Encryption Modular Semantic Task Framework) to secure semantic communication. This framework employs dynamic encryption during semantic extraction and feature transmission to hinder attackers from reconstructing data through eavesdropping, thus strengthening system privacy. To combat image communication task challenges, we propose a prototype adversarial collaborative alignment training approach enhanced by representation learning. This method extracts and encodes semantic features while using dynamic perturbation and robust optimization to improve system resilience against adversarial threats. The approach ensures reliable semantic communication in complex environments, maintaining performance while countering attacks using feature obfuscation, adversarial training, and representation learning. Experimental results demonstrate that our method surpasses existing techniques by more than 2% in resisting model inversion attacks on classification tasks. Visually, our method excels with minimal decipherable images for attackers. It also shows a 3% to 5% improvement in countering adversarial attacks on classification tasks.

ECAI Conference 2025 Conference Paper

Rethinking the Effect of LoRA in Foundation Models for Long-Tailed Recognition

• Haowei Liu
• Shijia Sun
• Liang Chen

Long-tailed recognition (LTR) has seen a surge in the level of attention it receives due to its practical value. Fine-tuning vision-language models (VLMs) has garnered significant attention among the various long-tailed approaches available, with foundation models thriving. While parameter-efficient fine-tuning (PEFT) methods such as adapter and visual prompt tuning (VPT) exhibit strong performance in long-tailed recognition, low-rank adaptation (LoRA), which is prominent in large language models (LLMs), fails to achieve comparable effectiveness in this context. To address the challenge, we introduce LotoRA, a groundbreaking long-tailed low-rank adaptation module. By leveraging diagonal blocks, LotoRA effectively enhances the rank while simultaneously reducing the number of parameters. This innovative approach overcomes the parameter limitations of traditional LoRA, enabling more efficient and targeted learning. Complementing this, we integrate Semantic Attention Pooling into the vision encoder and Semantic Prompt Embedding into the text encoder. These two components synergistically enhance the model’s capacity to represent tail classes by extracting more profound semantic features, effectively addressing the information deficiency often associated with tail categories in long-tailed datasets. Experimental results demonstrate that our method outperforms existing state-of-the-art (SOTA) approaches based on PEFT. Furthermore, our approach provides new insights into parameter-efficient adaptation for long-tailed recognition tasks with foundation models.

ICRA Conference 2025 Conference Paper

ROD: RGB-Only Fast and Efficient Off-Road Freespace Detection

• Tong Sun
• Hongliang Ye
• Jilin Mei
• Liang Chen
• Fangzhou Zhao
• Leiqiang Zong
• Yu Hu 0001

Off-road freespace detection is more challenging than on-road scenarios because of the blurred boundaries of traversable areas. Previous state-of-the-art (SOTA) methods employ multi-modal fusion of RGB images and LiDAR data. However, due to the significant increase in inference time when calculating surface normal maps from LiDAR data, multimodal methods are not suitable for real-time applications, particularly in real-world scenarios where higher FPS is required compared to slow navigation. This paper presents a novel RGB-only approach for off-road freespace detection, named ROD, eliminating the reliance on LiDAR data and its computational demands. Specifically, we utilize a pre-trained Vision Transformer (ViT) to extract rich features from RGB images. Additionally, we design a lightweight yet efficient decoder, which together improve both precision and inference speed. ROD establishes a new SOTA on ORFD and RELLIS-3D datasets, as well as an inference speed of 50 FPS, significantly outperforming prior models. Our code will be available at https://github.com/STLIFE97/offroad_roadseg.

NeurIPS Conference 2025 Conference Paper

SuperGPQA: Scaling LLM Evaluation across 285 Graduate Disciplines

• Xeron Du
• Yifan Yao
• Kaijing Ma
• Bingli Wang
• Tianyu Zheng
• Minghao Liu
• Yiming Liang
• Xiaolong Jin

Large language models (LLMs) have demonstrated remarkable proficiency in mainstream academic disciplines such as mathematics, physics, and computer science. However, human knowledge encompasses over 200 specialized disciplines, far exceeding the scope of existing benchmarks. The capabilities of LLMs in many of these specialized fields-particularly in light industry, agriculture, and service-oriented disciplines-remain inadequately evaluated. To address this gap, we present SuperGPQA, a comprehensive benchmark that evaluates graduate-level knowledge and reasoning capabilities across 285 disciplines. Our benchmark employs a novel Human-LLM collaborative filtering mechanism to eliminate trivial or ambiguous questions through iterative refinement based on both LLM responses and expert feedback. Our experimental results reveal significant room for improvement in the performance of current state-of-the-art LLMs across diverse knowledge domains (e. g. , the reasoning-focused model Gemini-2. 5-Pro achieved the highest accuracy of 63. 56% on SuperGPQA), highlighting the considerable gap between current model capabilities and artificial general intelligence. Additionally, we present comprehensive insights from our management of a large-scale annotation process, involving over 80 expert annotators and an interactive Human-LLM collaborative system, offering valuable methodological guidance for future research initiatives of comparable scope.

IROS Conference 2024 Conference Paper

A Safe and Efficient Timed-Elastic-Band Planner for Unstructured Environments

• Haoyu Xi
• Wei Li 0235
• Fangzhou Zhao
• Liang Chen
• Yu Hu 0001

In unstructured environments with complex obstacles and obscure road boundaries, the local planner faces more severe challenges in terms of safety and real-time performance. In order to fulfill these emerging requirements, we propose a novel Timed-Elastic-Band approach for unstructured environments, abbreviated as TEB-U. This approach incorporates a free space extraction optimization module for 2D occupancy grid maps, which efficiently transforms irregular free space boundaries into polygons and restrains robots within the boundaries. Moreover, a dynamic global point adjustment module is designed to adaptively correct the trajectory points obtained from the global planner, thereby enabling robots to travel along the centerline of free space and providing a better initial trajectory for subsequent modules. To reduce the computational cost, we replace the obstacle constraint of TEB with the boundary constraint in hyper-graph optimization. We evaluate our planner in three distinct scenarios, and the results show that TEB-U improves the average success rate by 21% and reduces the planning time by 23% compared to TEB in unstructured road, which demonstrates its safety and efficiency.

EAAI Journal 2024 Journal Article

A self-supervised contrastive change point detection method for industrial time series

• Xiangyu Bao
• Liang Chen
• Jingshu Zhong
• Dianliang Wu
• Yu Zheng

Manufacturing process monitoring is crucial to ensure production quality. This paper formulates the detection problem of abnormal changes in the manufacturing process as the change point detection (CPD) problem for the industrial temporal data. The premise of known data property and sufficient data annotations in existing CPD methods limits their application in the complex manufacturing process. Therefore, a self-supervised and non-parametric CPD method based on temporal trend-seasonal feature decomposition and contrastive learning (CoCPD) is proposed. CoCPD aims to solve CPD problem in an online manner. By bringing the representations of time series segments with similar properties in the feature space closer, our model can sensitively distinguish the change points that do not conform to either historical data distribution or temporal continuity. The proposed CoCPD is validated by a real-world body-in-white production case and compared with 10 state-of-the-art CPD methods. Overall, CoCPD achieves promising results by P r e c i s i o n 70. 6%, R e c a l l 68. 8%, and the mean absolute error ( M A E ) 8. 27. With the ability to rival the best offline baselines, CoCPD outperforms online baseline methods with improvements in P r e c i s i o n, R e c a l l and M A E by 14. 90%, 11. 93% and 43. 93%, respectively. Experiment results demonstrate that CoCPD can detect abnormal changes timely and accurately.

IJCAI Conference 2024 Conference Paper

Distribution-Independent Cell Type Identification for Single-Cell RNA-seq Data

• Yuyao Zhai
• Liang Chen
• Minghua Deng

Automatic cell type annotation aims to transfer the label knowledge from label-abundant reference data to label-scarce target data, which makes encouraging progress in single-cell RNA-seq data analysis. While previous works have focused on classifying close-set cells and detecting open-set cells during testing, it is still essential to be able to classify unknown cell types as human beings. Additionally, few efforts have been devoted to addressing the challenge of common long-tail dilemma in cell type annotation data. Therefore, in this paper, we propose an innovative distribution-independent universal cell type identification framework called scDET from the perspective of autonomously equilibrated dual-consultative contrastive learning. Our model can generate fine-grained predictions for both close-set and open-set cell types in a long-tailed open-world environment. scDET consists of a contrastive-learning branch and a pseudo-labeling branch, which work collaboratively to provide interactive supervision. Specifically, the contrastive-learning branch provides reliable distribution estimation to regularize the predictions of the pseudo-labeling branch, which in turn guides itself through self-balanced knowledge transfer and a designed novel soft contrastive loss. Extensive experimental results on various evaluation datasets demonstrate the superior performance of scDET over other state-of-the-art single-cell clustering and annotation methods.

NeurIPS Conference 2024 Conference Paper

LFME: A Simple Framework for Learning from Multiple Experts in Domain Generalization

• Liang Chen
• Yong Zhang
• Yibing Song
• Zhiqiang Shen
• Lingqiao Liu

Domain generalization (DG) methods aim to maintain good performance in an unseen target domain by using training data from multiple source domains. While success on certain occasions are observed, enhancing the baseline across most scenarios remains challenging. This work introduces a simple yet effective framework, dubbed learning from multiple experts (LFME), that aims to make the target model an expert in all source domains to improve DG. Specifically, besides learning the target model used in inference, LFME will also train multiple experts specialized in different domains, whose output probabilities provide professional guidance by simply regularizing the logit of the target model. Delving deep into the framework, we reveal that the introduced logit regularization term implicitly provides effects of enabling the target model to harness more information, and mining hard samples from the experts during training. Extensive experiments on benchmarks from different DG tasks demonstrate that LFME is consistently beneficial to the baseline and can achieve comparable performance to existing arts. Code is available at https: //github. com/liangchen527/LFME.

NeurIPS Conference 2024 Conference Paper

State Space Models on Temporal Graphs: A First-Principles Study

• Jintang Li
• Ruofan Wu
• Xinzhou Jin
• Boqun Ma
• Liang Chen
• Zibin Zheng

Over the past few years, research on deep graph learning has shifted from static graphs to temporal graphs in response to real-world complex systems that exhibit dynamic behaviors. In practice, temporal graphs are formalized as an ordered sequence of static graph snapshots observed at discrete time points. Sequence models such as RNNs or Transformers have long been the predominant backbone networks for modeling such temporal graphs. Yet, despite the promising results, RNNs struggle with long-range dependencies, while transformers are burdened by quadratic computational complexity. Recently, state space models (SSMs), which are framed as discretized representations of an underlying continuous-time linear dynamical system, have garnered substantial attention and achieved breakthrough advancements in independent sequence modeling. In this work, we undertake a principled investigation that extends SSM theory to temporal graphs by integrating structural information into the online approximation objective via the adoption of a Laplacian regularization term. The emergent continuous-time system introduces novel algorithmic challenges, thereby necessitating our development of GraphSSM, a graph state space model for modeling the dynamics of temporal graphs. Extensive experimental results demonstrate the effectiveness of our GraphSSM framework across various temporal graph benchmarks.

IROS Conference 2024 Conference Paper

TeFF: Tracking-enhanced Forgetting-free Few-shot 3D LiDAR Semantic Segmentation

• Junbao Zhou
• Jilin Mei
• Pengze Wu
• Liang Chen
• Fangzhou Zhao
• Xijun Zhao
• Yu Hu 0001

In autonomous driving, 3D LiDAR plays a crucial role in understanding the vehicle’s surroundings. However, the newly emerged, unannotated objects presents few-shot learning problem for semantic segmentation. This paper addresses the limitations of current few-shot semantic segmentation by exploiting the temporal continuity of LiDAR data. Employing a tracking model to generate pseudo-ground-truths from a sequence of LiDAR frames, our method significantly augments the dataset, enhancing the model’s ability to learn on novel classes. However, this approach introduces a data imbalance biased to novel data that presents a new challenge of catastrophic forgetting. To mitigate this, we incorporate LoRA, a technique that reduces the number of trainable parameters, thereby preserving the model’s performance on base classes while improving its adaptability to novel classes. This work represents a significant step forward in few-shot 3D LiDAR semantic segmentation for autonomous driving. Our code is available at https://github.com/BowmanChow/Track-no-forgetting.

NeurIPS Conference 2024 Conference Paper

UltraEdit: Instruction-based Fine-Grained Image Editing at Scale

• Haozhe Zhao
• Xiaojian Ma
• Liang Chen
• Shuzheng Si
• Rujie Wu
• Kaikai An
• Peiyu Yu
• Minjia Zhang

This paper presents UltraEdit, a large-scale (~ 4M editing samples), automatically generated dataset for instruction-based image editing. Our key idea is to address the drawbacks in existing image editing datasets like InstructPix2Pix and MagicBrush, and provide a systematic approach to producing massive and high-quality image editing samples: 1) UltraEdit includes more diverse editing instructions by combining LLM creativity and in-context editing examples by human raters; 2) UltraEdit is anchored on real images (photographs or artworks), which offers more diversity and less biases than those purely synthesized by text-to-image models; 3) UltraEdit supports region-based editing with high-quality, automatically produced region annotations. Our experiments show that canonical diffusion-based editing baselines trained on UltraEdit set new records on challenging MagicBrush and Emu-Edit benchmarks, respectively. Our analysis further confirms the crucial role of real image anchors and region-based editing data. The dataset, code, and models will be made public.

AAAI Conference 2024 Conference Paper

VQCNIR: Clearer Night Image Restoration with Vector-Quantized Codebook

• Wenbin Zou
• Hongxia Gao
• Tian Ye
• Liang Chen
• Weipeng Yang
• Shasha Huang
• Hongsheng Chen
• Sixiang Chen

Night photography often struggles with challenges like low light and blurring, stemming from dark environments and prolonged exposures. Current methods either disregard priors and directly fitting end-to-end networks, leading to inconsistent illumination, or rely on unreliable handcrafted priors to constrain the network, thereby bringing the greater error to the final result. We believe in the strength of data-driven high-quality priors and strive to offer a reliable and consistent prior, circumventing the restrictions of manual priors. In this paper, we propose Clearer Night Image Restoration with Vector-Quantized Codebook (VQCNIR) to achieve remarkable and consistent restoration outcomes on real-world and synthetic benchmarks. To ensure the faithful restoration of details and illumination, we propose the incorporation of two essential modules: the Adaptive Illumination Enhancement Module (AIEM) and the Deformable Bi-directional Cross-Attention (DBCA) module. The AIEM leverages the inter-channel correlation of features to dynamically maintain illumination consistency between degraded features and high-quality codebook features. Meanwhile, the DBCA module effectively integrates texture and structural information through bi-directional cross-attention and deformable convolution, resulting in enhanced fine-grained detail and structural fidelity across parallel decoders. Extensive experiments validate the remarkable benefits of VQCNIR in enhancing image quality under low-light conditions, showcasing its state-of-the-art performance on both synthetic and real-world datasets. The code is available at https://github.com/AlexZou14/VQCNIR.

JBHI Journal 2023 Journal Article

Deep Learning With Convolutional Neural Networks for Motor Brain-Computer Interfaces Based on Stereo-Electroencephalography (SEEG)

• Xiaolong Wu
• Shize Jiang
• Guangye Li
• Shengjie Liu
• Benjamin Metcalfe
• Liang Chen
• Dingguo Zhang

Objective: Deep learning based on convolutional neural networks (CNN) has achieved success in brain-computer interfaces (BCIs) using scalp electroencephalography (EEG). However, the interpretation of the so-called ‘black box’ method and its application in stereo-electroencephalography (SEEG)-based BCIs remain largely unknown. Therefore, in this paper, an evaluation is performed on the decoding performance of deep learning methods on SEEG signals. Methods: Thirty epilepsy patients were recruited, and a paradigm including five hand and forearm motion types was designed. Six methods, including filter bank common spatial pattern (FBCSP) and five deep learning methods (EEGNet, shallow and deep CNN, ResNet, and a deep CNN variant named STSCNN), were used to classify the SEEG data. Various experiments were conducted to investigate the effect of windowing, model structure, and the decoding process of ResNet and STSCNN. Results: The average classification accuracy for EEGNet, FBCSP, shallow CNN, deep CNN, STSCNN, and ResNet were 35 $\pm$ 6. 1%, 38 $\pm$ 4. 9%, 60 $\pm$ 3. 9%, 60 $\pm$ 3. 3%, 61 $\pm$ 3. 2%, and 63 $\pm$ 3. 1% respectively. Further analysis of the proposed method demonstrated clear separability between different classes in the spectral domain. Conclusion: ResNet and STSCNN achieved the first- and second-highest decoding accuracy, respectively. The STSCNN demonstrated that an extra spatial convolution layer was beneficial, and the decoding process can be partially interpreted from spatial and spectral perspectives. Significance: This study is the first to investigate the performance of deep learning on SEEG signals. In addition, this paper demonstrated that the so-called ‘black-box’ method can be partially interpreted.

AAAI Conference 2023 Conference Paper

Generalized Cell Type Annotation and Discovery for Single-Cell RNA-Seq Data

• Yuyao Zhai
• Liang Chen
• Minghua Deng

The rapid development of single-cell RNA sequencing (scRNA-seq) technology allows us to study gene expression heterogeneity at the cellular level. Cell annotation is the basis for subsequent downstream analysis in single-cell data mining. Existing methods rarely explore the fine-grained semantic knowledge of novel cell types absent from the reference data and usually susceptible to batch effects on the classification of seen cell types. Taking into consideration these limitations, this paper proposes a new and practical task called generalized cell type annotation and discovery for scRNA-seq data. In this task, cells of seen cell types are given class labels, while cells of novel cell types are given cluster labels instead of a unified “unassigned” label. To address this problem, we carefully design a comprehensive evaluation benchmark and propose a novel end-to-end algorithm framework called scGAD. Specifically, scGAD first builds the intrinsic correspondence across the reference and target data by retrieving the geometrically and semantically mutual nearest neighbors as anchor pairs. Then we introduce an anchor-based self-supervised learning module with a connectivity-aware attention mechanism to facilitate model prediction capability on unlabeled target data. To enhance the inter-type separation and intra-type compactness, we further propose a confidential prototypical self-supervised learning module to uncover the consensus category structure of the reference and target data. Extensive results on massive real datasets demonstrate the superiority of scGAD over various state-of-the-art clustering and annotation methods.

NeurIPS Conference 2023 Conference Paper

On the Pareto Front of Multilingual Neural Machine Translation

• Liang Chen
• Shuming Ma
• Dongdong Zhang
• Furu Wei
• Baobao Chang

In this work, we study how the performance of a given direction changes with its sampling ratio in Multilingual Neural Machine Translation (MNMT). By training over 200 multilingual models with various model sizes, data sizes, and language directions, we find it interesting that the performance of certain translation direction does not always improve with the increase of its weight in the multi-task optimization objective. Accordingly, scalarization method leads to a multitask trade-off front that deviates from the traditional Pareto front when there exists data imbalance in the training corpus, which poses a great challenge to improve the overall performance of all directions. Based on our observations, we propose the Double Power Law to predict the unique performance trade-off front in MNMT, which is robust across various languages, data adequacy, and the number of tasks. Finally, we formulate the sample ratio selection problem in MNMT as an optimization problem based on the Double Power Law. Extensive experiments show that it achieves better performance than temperature searching and gradient manipulation methods with only 1/5 to 1/2 of the total training budget. We release the code at https: //github. com/pkunlp-icler/ParetoMNMT for reproduction.

IJCAI Conference 2023 Conference Paper

Realistic Cell Type Annotation and Discovery for Single-cell RNA-seq Data

• Yuyao Zhai
• Liang Chen
• Minghua Deng

The rapid development of single-cell RNA sequencing (scRNA-seq) technologies allows us to explore tissue heterogeneity at the cellular level. Cell type annotation plays an essential role in the substantial downstream analysis of scRNA-seq data. Existing methods usually classify the novel cell types in target data as an “unassigned” group and rarely discover the fine-grained cell type structure among them. Besides, these methods carry risks, such as susceptibility to batch effect between reference and target data, thus further compromising of inherent discrimination of target data. Considering these limitations, here we propose a new and practical task called realistic cell type annotation and discovery for scRNA-seq data. In this task, cells from seen cell types are given class labels, while cells from novel cell types are given cluster labels. To tackle this problem, we propose an end-to-end algorithm framework called scPOT from the perspective of optimal transport (OT). Specifically, we first design an OT-based prototypical representation learning paradigm to encourage both global discriminations of clusters and local consistency of cells to uncover the intrinsic structure of target data. Then we propose an unbalanced OT-based partial alignment strategy with statistical filling to detect the cells from the seen cell types across reference and target data. Notably, scPOT also introduces an easy yet effective solution to automatically estimate the overall cell type number in target data. Extensive results on our carefully designed evaluation benchmarks demonstrate the superiority of scPOT over various state-of-the-art clustering and annotation methods.

IJCAI Conference 2023 Conference Paper

SAD: Semi-Supervised Anomaly Detection on Dynamic Graphs

• Sheng Tian
• Jihai Dong
• Jintang Li
• Wenlong Zhao
• Xiaolong Xu
• Baokun Wang
• Bowen Song
• Changhua Meng

Anomaly detection aims to distinguish abnormal instances that deviate significantly from the majority of benign ones. As instances that appear in the real world are naturally connected and can be represented with graphs, graph neural networks become increasingly popular in tackling the anomaly detection problem. Despite the promising results, research on anomaly detection has almost exclusively focused on static graphs while the mining of anomalous patterns from dynamic graphs is rarely studied but has significant application value. In addition, anomaly detection is typically tackled from semi-supervised perspectives due to the lack of sufficient labeled data. However, most proposed methods are limited to merely exploiting labeled data, leaving a large number of unlabeled samples unexplored. In this work, we present semi-supervised anomaly detection (SAD), an end-to-end framework for anomaly detection on dynamic graphs. By a combination of a time-equipped memory bank and a pseudo-label contrastive learning module, SAD is able to fully exploit the potential of large unlabeled samples and uncover underlying anomalies on evolving graph streams. Extensive experiments on four real-world datasets demonstrate that SAD efficiently discovers anomalies from dynamic graphs and outperforms existing advanced methods even when provided with only little labeled data.

AAAI Conference 2023 Conference Paper

Scaling Up Dynamic Graph Representation Learning via Spiking Neural Networks

• Jintang Li
• Zhouxin Yu
• Zulun Zhu
• Liang Chen
• Qi Yu
• Zibin Zheng
• Sheng Tian
• Ruofan Wu

Recent years have seen a surge in research on dynamic graph representation learning, which aims to model temporal graphs that are dynamic and evolving constantly over time. However, current work typically models graph dynamics with recurrent neural networks (RNNs), making them suffer seriously from computation and memory overheads on large temporal graphs. So far, scalability of dynamic graph representation learning on large temporal graphs remains one of the major challenges. In this paper, we present a scalable framework, namely SpikeNet, to efficiently capture the temporal and structural patterns of temporal graphs. We explore a new direction in that we can capture the evolving dynamics of temporal graphs with spiking neural networks (SNNs) instead of RNNs. As a low-power alternative to RNNs, SNNs explicitly model graph dynamics as spike trains of neuron populations and enable spike-based propagation in an efficient way. Experiments on three large real-world temporal graph datasets demonstrate that SpikeNet outperforms strong baselines on the temporal node classification task with lower computational costs. Particularly, SpikeNet generalizes to a large temporal graph (2.7M nodes and 13.9M edges) with significantly fewer parameters and computation overheads.

NeurIPS Conference 2023 Conference Paper

Towards Hybrid-grained Feature Interaction Selection for Deep Sparse Network

• Fuyuan Lyu
• Xing Tang
• Dugang Liu
• Chen Ma
• Weihong Luo
• Liang Chen
• Xiuqiang He
• Xue (Steve) Liu

Deep sparse networks are widely investigated as a neural network architecture for prediction tasks with high-dimensional sparse features, with which feature interaction selection is a critical component. While previous methods primarily focus on how to search feature interaction in a coarse-grained space, less attention has been given to a finer granularity. In this work, we introduce a hybrid-grained feature interaction selection approach that targets both feature field and feature value for deep sparse networks. To explore such expansive space, we propose a decomposed space which is calculated on the fly. We then develop a selection algorithm called OptFeature, which efficiently selects the feature interaction from both the feature field and the feature value simultaneously. Results from experiments on three large real-world benchmark datasets demonstrate that OptFeature performs well in terms of accuracy and efficiency. Additional studies support the feasibility of our method. All source code are publicly available\footnote{https: //anonymous. 4open. science/r/OptFeature-Anonymous}.

YNIMG Journal 2022 Journal Article

Assessing differential representation of hand movements in multiple domains using stereo-electroencephalographic recordings

• Guangye Li
• Shize Jiang
• Jianjun Meng
• Guohong Chai
• Zehan Wu
• Zhen Fan
• Jie Hu
• Xinjun Sheng

Invasive brain-computer interfaces (BCI) have made great progress in the reconstruction of fine hand movement parameters for paralyzed patients, where superficial measurement modalities including electrocorticography (ECoG) and micro-array recordings are mostly used. However, these recording techniques typically focus on the signals from the sensorimotor cortex, leaving subcortical regions and other cortical regions related to the movements largely unexplored. As an intracranial recording technique for the presurgical assessments of brain surgery, stereo-encephalography (SEEG) inserts depth electrodes containing multiple contacts into the brain and thus provides the unique opportunity for investigating movement-related neural representation throughout the brain. Although SEEG samples neural signals with high spatial-temporal resolutions, its potential of being used to build BCIs has just been realized recently, and the decoding of SEEG activity related to hand movements has not been comprehensively investigated yet. Here, we systematically evaluated the factors influencing the performance of movement decoding using SEEG signals recorded from 32 human subjects performing a visually-cued hand movement task. Our results suggest that multiple regions in both lateral and depth directions present significant neural selectivity to the task, whereas the sensorimotor area, including both precentral and postcentral cortex, carries the richest discriminative neural information for the decoding. The posterior parietal and prefrontal cortex contribute gradually less, but still rich sources for extracting movement parameters. The insula, temporal and occipital cortex also contains useful task-related information for decoding. Under the cortex layer, white matter presents decodable neural patterns but yields a lower accuracy (42.0 ± 0.8%) than the cortex on average (44.2 ± 0.8%, p<0.01). Notably, collectively using neural signals from multiple task-related areas can significantly enhance the movement decoding performance by 6.9% (p<0.01) on average compared to using a single region. Among the different spectral components of SEEG activity, the high gamma and delta bands offer the most informative features for hand movements reconstruction. Additionally, the phase-amplitude coupling strength between these two frequency ranges correlates positively with the performance of movement decoding. In the temporal domain, maximum decoding accuracy is first reached around 2 s after the onset of movement commands. In sum, this study provides valuable insights for the future motor BCIs design employing both SEEG recordings and other recording modalities.

AAAI Conference 2022 Conference Paper

Evidential Neighborhood Contrastive Learning for Universal Domain Adaptation

• Liang Chen
• Yihang Lou
• Jianzhong He
• Tao Bai
• Minghua Deng

Universal domain adaptation (UniDA) aims to transfer the knowledge learned from a labeled source domain to an unlabeled target domain without any constraints on the label sets. However, domain shift and category shift make UniDA extremely challenging, mainly attributed to the requirement of identifying both shared “known” samples and private “unknown” samples. Previous methods barely exploit the intrinsic manifold structure relationship between two domains for feature alignment, and they rely on the softmax-based scores with class competition nature to detect underlying “unknown” samples. Therefore, in this paper, we propose a novel evidenTial Neighborhood conTrastive learning framework called TNT to address these issues. Specifically, TNT first proposes a new domain alignment principle: semantically consistent samples should be geometrically adjacent to each other, whether within or across domains. From this criterion, a cross domain multi-sample contrastive loss based on mutual nearest neighbors is designed to achieve common category matching and private category separation. Second, toward accurate “unknown” sample detection, TNT introduces a class competition-free uncertainty score from the perspective of evidential deep learning. Instead of setting a single threshold, TNT learns a category-aware heterogeneous threshold vector to reject diverse “unknown” samples. Extensive experiments on three benchmarks demonstrate that TNT significantly outperforms previous state-of-the-art UniDA methods.

NeurIPS Conference 2022 Conference Paper

Large-batch Optimization for Dense Visual Predictions: Training Faster R-CNN in 4.2 Minutes

• Zeyue Xue
• Jianming Liang
• Guanglu Song
• Zhuofan Zong
• Liang Chen
• Yu Liu
• Ping Luo

Training a large-scale deep neural network in a large-scale dataset is challenging and time-consuming. The recent breakthrough of large-batch optimization is a promising way to tackle this challenge. However, although the current advanced algorithms such as LARS and LAMB succeed in classification models, the complicated pipelines of dense visual predictions such as object detection and segmentation still suffer from the heavy performance drop in the large-batch training regime. To address this challenge, we propose a simple yet effective algorithm, named Adaptive Gradient Variance Modulator (AGVM), which can train dense visual predictors with very large batch size, enabling several benefits more appealing than prior arts. Firstly, AGVM can align the gradient variances between different modules in the dense visual predictors, such as backbone, feature pyramid network (FPN), detection, and segmentation heads. We show that training with a large batch size can fail with the gradient variances misaligned among them, which is a phenomenon primarily overlooked in previous work. Secondly, AGVM is a plug-and-play module that generalizes well to many different architectures (e. g. , CNNs and Transformers) and different tasks (e. g. , object detection, instance segmentation, semantic segmentation, and panoptic segmentation). It is also compatible with different optimizers (e. g. , SGD and AdamW). Thirdly, a theoretical analysis of AGVM is provided. Extensive experiments on the COCO and ADE20K datasets demonstrate the superiority of AGVM. For example, AGVM demonstrates more stable generalization performance than prior arts under extremely large batch size (i. e. , 10k). AGVM can train Faster R-CNN+ResNet50 in 4. 2 minutes without losing performance. It enables training an object detector with one billion parameters in just 3. 5 hours, reducing the training time by 20. 9×, whilst achieving 62. 2 mAP on COCO. The deliverables will be released at https: //github. com/Sense-X/AGVM.

AAAI Conference 2022 Conference Paper

Mutual Nearest Neighbor Contrast and Hybrid Prototype Self-Training for Universal Domain Adaptation

• Liang Chen
• Qianjin Du
• Yihang Lou
• Jianzhong He
• Tao Bai
• Minghua Deng

Universal domain adaptation (UniDA) aims to transfer knowledge learned from a labeled source domain to an unlabeled target domain under domain shift and category shift. Without prior category overlap information, it is challenging to simultaneously align the common categories between two domains and separate their respective private categories. Additionally, previous studies utilize the source classifier’s prediction to obtain various known labels and one generic “unknown” label of target samples. However, overreliance on learned classifier knowledge is inevitably biased to source data, ignoring the intrinsic structure of target domain. Therefore, in this paper, we propose a novel two-stage UniDA framework called MATHS based on the principle of Mutual neArest neighbor conTrast and Hybrid prototype diScrimination. In the first stage, we design an efficient mutual nearest neighbor contrastive learning scheme to achieve feature alignment, which exploits the instance-level affinity relationship to uncover the intrinsic structure of two domains. We introduce a bimodality hypothesis for the maximum discriminative probability distribution to detect the possible target private samples, and present a data-based statistical approach to separate the common and private categories. In the second stage, to obtain more reliable label predictions, we propose an incremental pseudo-classifier for target data only, which is driven by the hybrid representative prototypes. A confidence-guided prototype contrastive loss is designed to optimize the category allocation uncertainty via a selftraining mechanism. Extensive experiments on three benchmarks demonstrate that MATHS outperforms previous stateof-the-arts on most UniDA settings.

AAAI Conference 2022 Conference Paper

Neighborhood Consensus Contrastive Learning for Backward-Compatible Representation

• Shengsen Wu
• Liang Chen
• Yihang Lou
• Yan Bai
• Tao Bai
• Minghua Deng
• Ling-Yu Duan

In object re-identification (ReID), the development of deep learning techniques often involves model updates and deployment. It is unbearable to re-embedding and re-index with the system suspended when deploying new models. Therefore, backward-compatible representation is proposed to enable “new” features to be compared with “old” features directly, which means that the database is active when there are both “new” and “old” features in it. Thus we can scroll-refresh the database or even do nothing on the database to update. The existing backward-compatible methods either require a strong overlap between old and new training data or simply conduct constraints at the instance level. Thus they are difficult in handling complicated cluster structures and are limited in eliminating the impact of outliers in old embeddings, resulting in a risk of damaging the discriminative capability of new features. In this work, we propose a Neighborhood Consensus Contrastive Learning (NCCL) method. With no assumptions about the new training data, we estimate the subcluster structures of old embeddings. A new embedding is constrained with multiple old embeddings in both embedding space and discrimination space at the sub-class level. The effect of outliers diminished, as the multiple samples serve as “mean teachers”. Besides, we propose a scheme to filter the old embeddings with low credibility, further improving the compatibility robustness. Our method ensures the compatibility without impairing the accuracy of the new model. It can even improve the new model’s accuracy in most scenarios.

NeurIPS Conference 2022 Conference Paper

OST: Improving Generalization of DeepFake Detection via One-Shot Test-Time Training

• Liang Chen
• Yong Zhang
• Yibing Song
• Jue Wang
• Lingqiao Liu

State-of-the-art deepfake detectors perform well in identifying forgeries when they are evaluated on a test set similar to the training set, but struggle to maintain good performance when the test forgeries exhibit different characteristics from the training images e. g. , forgeries are created by unseen deepfake methods. Such a weak generalization capability hinders the applicability of deepfake detectors. In this paper, we introduce a new learning paradigm specially designed for the generalizable deepfake detection task. Our key idea is to construct a test-sample-specific auxiliary task to update the model before applying it to the sample. Specifically, we synthesize pseudo-training samples from each test image and create a test-time training objective to update the model. Moreover, we proposed to leverage meta-learning to ensure that a fast single-step test-time gradient descent, dubbed one-shot test-time training (OST), can be sufficient for good deepfake detection performance. Extensive results across several benchmark datasets demonstrate that our approach performs favorably against existing arts in terms of generalization to unseen data and robustness to different post-processing steps.

IJCAI Conference 2022 Conference Paper

Spiking Graph Convolutional Networks

• Zulun Zhu
• Jiaying Peng
• Jintang Li
• Liang Chen
• Qi Yu
• Siqiang Luo

Graph Convolutional Networks (GCNs) achieve an impressive performance due to the remarkable representation ability in learning the graph information. However, GCNs, when implemented on a deep network, require expensive computation power, making them difficult to be deployed on battery-powered devices. In contrast, Spiking Neural Networks (SNNs), which perform a bio-fidelity inference process, offer an energy-efficient neural architecture. In this work, we propose SpikingGCN, an end-to-end framework that aims to integrate the embedding of GCNs with the biofidelity characteristics of SNNs. The original graph data are encoded into spike trains based on the incorporation of graph convolution. We further model biological information processing by utilizing a fully connected layer combined with neuron nodes. In a wide range of scenarios (e. g. , citation networks, image graph classification, and recommender systems), our experimental results show that the proposed method could gain competitive performance against state-of-the-art approaches. Furthermore, we show that SpikingGCN on a neuromorphic chip can bring a clear advantage of energy efficiency into graph data analysis, which demonstrates its great potential to construct environment-friendly machine learning models.

YNIMG Journal 2021 Journal Article

Prestimulus dynamics blend with the stimulus in neural variability quenching

• Annemarie Wolff
• Liang Chen
• Shankar Tumati
• Mehrshad Golesorkhi
• Javier Gomez-Pilar
• Jie Hu
• Shize Jiang
• Ying Mao

Neural responses to the same stimulus show significant variability over trials, with this variability typically reduced (quenched) after a stimulus is presented. This trial-to-trial variability (TTV) has been much studied, however how this neural variability quenching is influenced by the ongoing dynamics of the prestimulus period is unknown. Utilizing a human intracranial stereo-electroencephalography (sEEG) data set, we investigate how prestimulus dynamics, as operationalized by standard deviation (SD), shapes poststimulus activity through trial-to-trial variability (TTV). We first observed greater poststimulus variability quenching in those real trials exhibiting high prestimulus variability as observed in all frequency bands. Next, we found that the relative effect of the stimulus was higher in the later (300-600ms) than the earlier (0-300ms) poststimulus period. Lastly, we replicate our findings in a separate EEG dataset and extend them by finding that trials with high prestimulus variability in the theta and alpha bands had faster reaction times. Together, our results demonstrate that stimulus-related activity, including its variability, is a blend of two factors: 1) the effects of the external stimulus itself, and 2) the effects of the ongoing dynamics spilling over from the prestimulus period - the state at stimulus onset - with the second dwarfing the influence of the first.

AAAI Conference 2021 Conference Paper

Reinforcement Learning with a Disentangled Universal Value Function for Item Recommendation

• Kai Wang
• Zhene Zou
• Qilin Deng
• Jianrong Tao
• Runze Wu
• Changjie Fan
• Liang Chen
• Peng Cui

In recent years, there are great interests as well as challenges in applying reinforcement learning (RL) to recommendation systems (RS). In this paper, we summarize three key practical challenges of large-scale RL-based recommender systems: massive state and action spaces, high-variance environment, and the unspecific reward setting in recommendation. All these problems remain largely unexplored in the existing literature and make the application of RL challenging. We develop a model-based reinforcement learning framework, called GoalRec. Inspired by the ideas of world model (model-based), value function estimation (model-free), and goal-based RL, a novel disentangled universal value function designed for item recommendation is proposed. It can generalize to various goals that the recommender may have, and disentangle the stochastic environmental dynamics and high-variance reward signals accordingly. As a part of the value function, free from the sparse and high-variance reward signals, a high-capacity reward-independent world model is trained to simulate complex environmental dynamics under a certain goal. Based on the predicted environmental dynamics, the disentangled universal value function is related to the user’s future trajectory instead of a monolithic state and a scalar reward. We demonstrate the superiority of GoalRec over previous approaches in terms of the above three practical challenges in a series of simulations and a real application.

IJCAI Conference 2021 Conference Paper

Secure Deep Graph Generation with Link Differential Privacy

• Carl Yang
• Haonan Wang
• Ke Zhang
• Liang Chen
• Lichao Sun

Many data mining and analytical tasks rely on the abstraction of networks (graphs) to summarize relational structures among individuals (nodes). Since relational data are often sensitive, we aim to seek effective approaches to generate utility-preserved yet privacy-protected structured data. In this paper, we leverage the differential privacy (DP) framework to formulate and enforce rigorous privacy constraints on deep graph generation models, with a focus on edge-DP to guarantee individual link privacy. In particular, we enforce edge-DP by injecting designated noise to the gradients of a link reconstruction based graph generation model, while ensuring data utility by improving structure learning with structure-oriented graph discrimination. Extensive experiments on two real-world network datasets show that our proposed DPGGAN model is able to generate graphs with effectively preserved global structure and rigorously protected individual link privacy.

IJCAI Conference 2021 Conference Paper

Understanding Structural Vulnerability in Graph Convolutional Networks

• Liang Chen
• Jintang Li
• Qibiao Peng
• Yang Liu
• Zibin Zheng
• Carl Yang

Recent studies have shown that Graph Convolutional Networks (GCNs) are vulnerable to adversarial attacks on the graph structure. Although multiple works have been proposed to improve their robustness against such structural adversarial attacks, the reasons for the success of the attacks remain unclear. In this work, we theoretically and empirically demonstrate that structural adversarial examples can be attributed to the non-robust aggregation scheme (i. e. , the weighted mean) of GCNs. Specifically, our analysis takes advantage of the breakdown point which can quantitatively measure the robustness of aggregation schemes. The key insight is that weighted mean, as the basic design of GCNs, has a low breakdown point and its output can be dramatically changed by injecting a single edge. We show that adopting the aggregation scheme with a high breakdown point (e. g. , median or trimmed mean) could significantly enhance the robustness of GCNs against structural attacks. Extensive experiments on four real-world datasets demonstrate that such a simple but effective method achieves the best robustness performance compared to state-of-the-art models.

IJCAI Conference 2019 Conference Paper

Matching User with Item Set: Collaborative Bundle Recommendation with Deep Attention Network

• Liang Chen
• Yang Liu
• Xiangnan He
• Lianli Gao
• Zibin Zheng

Most recommendation research has been concentrated on recommending single items to users, such as the considerable work on collaborative filtering that models the interaction between a user and an item. However, in many real-world scenarios, the platform needs to show users a set of items, e. g. , the marketing strategy that offers multiple items for sale as one bundle. In this work, we consider recommending a set of items to a user, i. e. , the Bundle Recommendation task, which concerns the interaction modeling between a user and a set of items. We contribute a neural network solution named DAM, short for Deep Attentive Multi-Task model, which is featured with two special designs: 1) We design a factorized attention network to aggregate the item embeddings in a bundle to obtain the bundle's representation; 2) We jointly model user-bundle interactions and user-item interactions in a multi-task manner to alleviate the scarcity of user-bundle interactions. Extensive experiments on a real-world dataset show that DAM outperforms the state-of-the-art solution, verifying the effectiveness of our attention design and multi-task learning in DAM.

YNICL Journal 2019 Journal Article

Physiological significance of R-fMRI indices: Can functional metrics differentiate structural lesions (brain tumors)?

• Zhen Fan
• Xiao Chen
• Zeng-Xin Qi
• Le Li
• Bin Lu
• Cong-Lin Jiang
• Ren-Qing Zhu
• Chao-Gan Yan

Resting-state functional MRI (R-fMRI) research has recently entered the era of "big data", however, few studies have provided a rigorous validation of the physiological underpinnings of R-fMRI indices. Although studies have reported that various neuropsychiatric disorders exhibit abnormalities in R-fMRI measures, these "biomarkers" have not been validated in differentiating structural lesions (brain tumors) as a concept proof. We enrolled 60 patients with intracranial tumors located in the unilateral cranialcavity and 60 matched normal controls to test whether R-fMRI indices can differentiate tumors, which represents a prerequisite for adapting such indices as biomarkers for neuropsychiatric disorders. Common R-fMRI indices of tumors and their counterpart control regions, which were defined as the contralateral normal areas (for amplitude of low frequency fluctuations (ALFF), fractional ALFF (fALFF), regional homogeneity (ReHo) and degree centrality (DC)) and ipsilateral regions surrounding the tumors (for voxel-mirrored homotopic connectivity (VMHC)), were comprehensively assessed. According to robust paired t-tests with a Bonferroni correction, only VMHC (Fisher's r-to-z transformed) could successfully differentiate substantial tumors from their counterpart normal regions in patients. Furthermore, ALFF and DC were not able to differentiate tumor from normal unless Z-standardization was employed. To validate the lower power of the between-subject design compared to the within-subject design, each metric was calculated in a matched control group, and robust two-sample t-tests were used to compare the patient tumors and the normal controls at the same place. Similarly, only VMHC succeeded in differentiating significant differences between tumors and the sham tumor areas of normal controls. This study tested the premise of R-fMRI biomarkers for differentiating lesions, and brings a new understanding to physical significance of the Z-standardization.

YNIMG Journal 2018 Journal Article

Optimal referencing for stereo-electroencephalographic (SEEG) recordings

• Guangye Li
• Shize Jiang
• Sivylla E. Paraskevopoulou
• Meng Wang
• Yang Xu
• Zehan Wu
• Liang Chen
• Dingguo Zhang

Stereo-electroencephalography (SEEG) is an intracranial recording technique in which depth electrodes are inserted in the brain as part of presurgical assessments for invasive brain surgery. SEEG recordings can tap into neural signals across the entire brain and thereby sample both cortical and subcortical sites. However, even though signal referencing is important for proper assessment of SEEG signals, no previous study has comprehensively evaluated the optimal referencing method for SEEG. In our study, we recorded SEEG data from 15 human subjects during a motor task, referencing them against the average of two white matter contacts (monopolar reference). We then subjected these signals to 5 different re-referencing approaches: common average reference (CAR), gray-white matter reference (GWR), electrode shaft reference (ESR), bipolar reference, and Laplacian reference. The results from three different signal quality metrics suggest the use of the Laplacian re-reference for study of local population-level activity and low-frequency oscillatory activity.

YNICL Journal 2017 Journal Article

Fully automatic acute ischemic lesion segmentation in DWI using convolutional neural networks

• Liang Chen
• Paul Bentley
• Daniel Rueckert

Stroke is an acute cerebral vascular disease, which is likely to cause long-term disabilities and death. Acute ischemic lesions occur in most stroke patients. These lesions are treatable under accurate diagnosis and treatments. Although diffusion-weighted MR imaging (DWI) is sensitive to these lesions, localizing and quantifying them manually is costly and challenging for clinicians. In this paper, we propose a novel framework to automatically segment stroke lesions in DWI. Our framework consists of two convolutional neural networks (CNNs): one is an ensemble of two DeconvNets (Noh et al., 2015), which is the EDD Net; the second CNN is the multi-scale convolutional label evaluation net (MUSCLE Net), which aims to evaluate the lesions detected by the EDD Net in order to remove potential false positives. To the best of our knowledge, it is the first attempt to solve this problem and using both CNNs achieves very good results. Furthermore, we study the network architectures and key configurations in detail to ensure the best performance. It is validated on a large dataset comprising clinical acquired DW images from 741 subjects. A mean accuracy of Dice coefficient obtained is 0.67 in total. The mean Dice scores based on subjects with only small and large lesions are 0.61 and 0.83, respectively. The lesion detection rate achieved is 0.94.

AAMAS Conference 2016 Conference Paper

Decision Theoretic Norm-Governed Planning (Extended Abstract)

• Luca Gasparini
• Timothy J. Norman
• Martin J. Kollingbaum
• Liang Chen

We propose Normative Dec-POMDPs, a model of collective decision making in the presence of complex norms, with violations of norms classified according to their relative severity. We extend the PBPG algorithm in order to solve Normative Dec-POMDPs and propose a heuristic that improves its scalability without affecting the policy quality.

AIJ Journal 2005 Journal Article

A general stability analysis on regional and national voting schemes against noise—why is an electoral college more stable than a direct popular election?

• Liang Chen
• Naoyuki Tokuda

By discarding the previous restrictive weak average distribution assumption on region sizes, we have developed a new general probabilistic model on the regional voting (known as “direct popular voting” in political science) and the national voting (typically, the electoral college), where we regard the percentage of a candidate's supporters in the nation as the probability of a voter voting for the candidate. Our analysis demonstrates that the regional voting is always more stable than the national voting, and that the stability margin of the regional voting always increases as the size of such partitioned regions decreases down to a certain critical value of region size, beyond which the stability margin starts to decrease, asymptoting to a national voting level where the size of the partitioned regions approaches the unit of voting cell so that the improved stability of the regional voting by localizing the effects of noise into a restricted number of smaller effective areas will not be effective. Our stability analysis remains valid over the entire range in size of the partitioned regions for regional voting. We show that the regional voting asymptotes to the national voting in two extreme limiting cases, when the region size decreases to a voting cell size and when the region size increases to the size of the nation.

AIJ Journal 2003 Journal Article

Robustness of regional matching scheme over global matching scheme

• Liang Chen
• Naoyuki Tokuda

Analyzing the effect of concentrated noise on a typical decision-making process of a simplified two-candidate voting model, we have demonstrated that a local approach using a regional matching process is more robust and stable than a direct approach using a global matching process, by establishing that the former is capable of accommodating a higher level of noise than the latter before the result of the decision overturns. To extend the theory to imagery analysis, we pose a conjecture that our conclusion on the robustness of the regional matching processes remains valid not only for the simpler vote counting schemes but also for practically more important decision-making schemes in image analysis which involve dimension-reducing transforms or other features extraction processes such as principal component analysis or Gablor transforms. Two convincing experimental verifications are provided, supporting not only the theory by a white-black flag recognition problem on a pixel-by-pixel basis, but also the validity of the conjecture by a facial recognition problem in the presence of localized noise typically represented by clutter or occlusion in imagery.

MFCS Conference 1993 Conference Paper

A Model for Real-Time Process Algebras (Extended Abstract)

• Liang Chen