Author name cluster

Cheng Cheng

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

19 papers

2 author rows

YNIMG Journal 2026 Journal Article

Emotion processing models in individuals with different levels of social anxiety based on the crowd emotion perception task

Cheng Cheng
Shuaixia Li
Yuanyuan Yang
Chunlin Wu
Yajing Si
Weiqi He

Previous studies on dyadic face perception have shown that individuals with high social anxiety (HSA) exhibit a strong attentional bias toward threat-related faces. However, it remains unclear whether this bias persists when multiple faces appear simultaneously. The present study combined an average emotion judgment task with event-related potentials (ERP) and functional brain network analysis to examine the temporal dynamics of crowd facial expression processing and the influence of social anxiety. Behaviorally, individuals with HSA identified crowd emotional expressions more accurately than individuals with low social anxiety (LSA). ERP results revealed enhanced late positive potential (LPP) amplitudes to angry versus happy crowd faces only in the HSA group, indicating sustained attention to social threat. In contrast, Dual Laplacian Granger Causality Analysis (DLap-GCA) showed stronger frontal-parietal and frontal-occipital connectivity in the LSA group, suggesting more efficient integration of regulatory, attentional, and visual processes during crowd emotion perception. A graph neural network trained on these connectivity features achieved an average accuracy of 87.14% in discriminating anxiety groups. These findings indicate that social anxiety modulates crowd emotion processing through distinct mechanisms: HSA is characterized by amplified threat sensitivity, whereas LSA relies on stronger regulatory network coordination that supports stable and efficient emotional integration in socially threatening contexts.

Details DOI

JBHI Journal 2026 Journal Article

MSDB-Mam: Dual-Branch Mamba Network With Multi-Scale Features for EEG-Based Depression Detection

Cheng Cheng
Wenzhe Liu
Peiyang Li
Ziyu Jia
Wenbo Luo

Depression is a widespread mental illness, and EEG-based detection offers a non-invasive means to reflect brain activity. However, the inherent complexity of EEG signals across spatial, temporal, and spectral domains poses major challenges for accurate detection, since brain activities vary dynamically over time, differ across brain regions, and are distributed across multiple frequency bands. To this end, we propose a novel Multi-Scale Dual-Branch Mamba network (MSDB-Mam) that efficiently extracts and fuses multi-dimensional EEG features. Specifically, a Multi-Scale Convolution (MSC) module captures diverse patterns along temporal and spatial-temporal axes using different kernel sizes. An Adaptive Reallocation (AR) unit dynamically adjusts feature weights to highlight informative patterns. To capture long-range and cross-domain dependencies, we introduce a Dual-Branch Mamba (DB-Mam) architecture, consisting of a Temporal–Spectral Mamba (TS-Mam) branch for modeling temporal–frequency correlations and a Spatial–Temporal–Spectral Mamba (STS-Mam) branch for learning richer interactions across spatial, temporal, and frequency. The features from both branches are subsequently fused to form a comprehensive and expressive EEG representation. Experiments on MODMA and PRED+CT datasets show that our method achieves 96. 58% and 96. 66% accuracy, respectively, surpassing existing approaches and demonstrating its effectiveness in EEG-based depression detection.

Details DOI

ICML Conference 2025 Conference Paper

CogMath: Assessing LLMs' Authentic Mathematical Ability from a Human Cognitive Perspective

Jiayu Liu 0001
Zhenya Huang
Wei Dai
Cheng Cheng
Jinze Wu
Jing Sha
Song Li
Qi Liu 0003

Although large language models (LLMs) show promise in solving complex mathematical tasks, existing evaluation paradigms rely solely on a coarse measure of overall answer accuracy, which are insufficient for assessing their authentic capabilities. In this paper, we propose CogMath, which comprehensively assesses LLMs’ mathematical abilities through the lens of human cognition. Specifically, inspired by psychological theories, CogMath formalizes human reasoning process into 3 stages: problem comprehension, problem solving, and solution summarization. Within these stages, we investigate perspectives such as numerical calculation, knowledge, and counterfactuals, and design a total of 9 fine-grained evaluation dimensions. In each dimension, we develop an “ Inquiry - Judge - Reference ” multi-agent system to generate inquiries that assess LLMs’ mastery from this dimension. An LLM is considered to truly master a problem only when excelling in all inquiries from the 9 dimensions. By applying CogMath on three benchmarks, we reveal that the mathematical capabilities of 7 mainstream LLMs are overestimated by 30%-40%. Moreover, we locate their strengths and weaknesses across specific stages/dimensions, offering in-depth insights to further enhance their reasoning abilities.

Details

EAAI Journal 2025 Journal Article

Knowledge-data synergy enabling zero-shot composite fault diagnosis in sucker-rod pumping systems

Xin-yan Wang
Li-ming Zhang
Kai Zhang
Cheng Cheng

Zero-Shot Learning Accurate and comprehensive fault diagnosis is critical to ensuring production efficiency and extending equipment lifespan in oilfield operations. However, most existing diagnostic methods for sucker rod pumping systems are limited to single-fault identification, while the scarcity and high acquisition cost of compound fault samples hinder the application of supervised learning. To address these challenges, this paper proposes a zero-shot compound fault diagnosis framework that integrates domain knowledge with data-driven representations, enabling the identification of unseen fault using only single-fault samples. First, a unified semantic space is constructed by jointly embedding text-based fault descriptions and capsule-encoded load features. Then, a semantics-guided weakly supervised attribute composition strategy is introduced to enhance the completeness and discriminability of the semantic space. Finally, a bidirectional contrastive learning mechanism is established between visual encodings and semantic representations, and an adaptive multi-task loss weighting strategy is employed to optimize the overall framework efficiently. Experiments on real-world oilfield data demonstrate that the proposed method achieves an F1-score of 86. 15 % under the zero-shot setting and 69. 08 % under the generalized zero-shot setting, offering an effective and scalable solution for compound fault diagnosis in oil wells.

Details DOI

ICML Conference 2025 Conference Paper

LoRA-Gen: Specializing Large Language Model via Online LoRA Generation

Yicheng Xiao
Lin Song 0002
Rui Yan 0001
Cheng Cheng
Yixiao Ge
Xiu Li 0001
Ying Shan

Recent advances have highlighted the benefits of scaling language models to enhance performance across a wide range of NLP tasks. However, these approaches still face limitations in effectiveness and efficiency when applied to domain-specific tasks, particularly for small edge-side models. We propose the LoRA-Gen framework, which utilizes a large cloud-side model to generate LoRA parameters for edge-side models based on task descriptions. By employing the reparameterization technique, we merge the LoRA parameters into the edge-side model to achieve flexible specialization. Our method facilitates knowledge transfer between models while significantly improving the inference efficiency of the specialized model by reducing the input context length. Without specialized training, LoRA-Gen outperforms conventional LoRA fine-tuning, which achieves competitive accuracy and a 2. 1x speedup with TinyLLaMA-1. 1B in reasoning tasks. Besides, our method delivers a compress ratio of 10. 1x with Gemma-2B on intelligent agent tasks.

Details

AAAI Conference 2025 Conference Paper

Multi-Perspective Consolidation Enhanced Cognitive Diagnosis via Conditional Diffusion Model

Guanhao Zhao
Zhenya Huang
Cheng Cheng
Yan Zhuang
Qingyang Mao
Xin Li
Shijin Wang
Enhong Chen

Cognitive diagnosis, which assesses the learners' competence from learners' interaction logs, plays a vital role in education. It provides a crucial reference for gauging learners' proficiency levels and tailoring future learning activities accordingly. Researchers have proposed numerous cognitive diagnosis models to address this task. Despite their success, these models continue to face the ill-posed problem because of the information loss caused by under-expressive interaction function and incomplete observations. In this paper, we address these challenges by proposing a novel cognitive diagnosis model, DMC-CDM, based on the theoretical premise that cognitive states can be captured with minimal information loss by maximizing the mutual information between observed and potential observations. Specifically, DMC-CDM incorporates a semantic extractor to provide a comprehensive semantic understanding of learners' interaction logs, thereby enhancing current collaborative-based cognitive state representations. It then consolidates multi-perspective observations to capture precise cognitive states by maximizing mutual information between these observations. We conducted extensive experiments on three datasets, and the experimental results demonstrate that our proposed model is both effective and beneficial for downstream applications in education.

PDF Details DOI

EAAI Journal 2025 Journal Article

Multi-scale latent space sequential fusion of images for pose estimation of underground tunneling machinery

Hongzhuang Wu
Cheng Cheng
Deyi Zhang
Hongzhi Zhou

Recently, the underground unmanned tunneling technology has received increasing attention, in which the pose estimation of the tunneling machinery occupies a prominent position. However, the autonomous positioning of the tunneling machinery presents significant challenges due to the harsh underground environments and the extremely complicated working conditions, which severely restricts the development and application of underground intelligent tunneling technology. Therefore, this paper investigates a machine vision and deep learning based pose estimation method for the underground tunneling machinery. Considering that both effective features and interference features of time-series images exhibit spatiotemporal correlations across multiple scales, we propose a multi-scale latent space sequential fusion (MSLSSF) model to integrate information from time-sequence images and further estimate the pose variables of the tunneling machine. The multi-scale variational autoencoder (MSVAE) is employed in the proposed model to obtain multi-scale latent space representations for each image in the time sequence. The proposed MSLSSF strategy consists of two steps. The first step involves fusing the latent representations of the time-series images at each scale using long short-term memory (LSTM) models, as specific correlations may exist in the representations of the sequential images across various scales. The second step employs an attention mechanism to adaptively fuse the results from the first step across all scales and timesteps. The presented information fusion methods provide a way for the model to take full advantage of the multi-scale features of time-sequence images. Experiments are conducted on our custom-made tunneling machine visual localization dataset, where the proposed MSLSSF based pose estimation method outperforms the advanced comparison methods in both accuracy and robustness, thereby validating the efficacy of the proposed approach. Additionally, the feasibility of the MSLSSF strategy is confirmed through the visualization results of the proposed model.

Details DOI

EAAI Journal 2025 Journal Article

Multifunctional health status assessment based on decentralized federated temporal domain adaptation for rotating machinery

Wei Zhou
Yong Zhang
Zuowei Ping
Cheng Cheng
Jiahua Sun

Based on the training with the large amounts of labeled data, intelligent data-driven approaches are able to assess the health status of rotating machinery. Traditional methods usually lack the multifunctionality required for health status assessment (HSA), resulting in incomplete assessment results. Furthermore, considering the challenge of machinery data silos, existing methods use collaborative model training solutions with multiple users, which place high demands on data privacy protection due to conflict of interests. To tackle this issue, a multifunctional HSA method based on decentralized federated temporal domain adaptation is proposed in this paper. First, a novel multifunctional HSA framework is designed for comprehensive assessment of rotating machinery, which performs synchronous health stage division and recognition, remaining useful life prediction, and reliability evaluation. Then, a temporal adversarial domain adaptation model is proposed to align temporal feature distributions through adversarial training between cross-domain encoders and a discriminator, both integrated with channel attention to enhance temporal feature extraction capability. Meanwhile, data privacy is ensured through decentralized federated learning involving multiple users. Experimental results show that the proposed method exhibits better generalization and stability, achieving an average error of 0. 087 in dual prediction functions and an accuracy of 0. 94 in recognition functionality.

Details DOI

EAAI Journal 2025 Journal Article

Reliable federated learning based on delayed gradient aggregation for intelligent connected vehicles

Zhigang Yang
Cheng Cheng
Zixuan Li
Ruyan Wang
Xuhua Zhang

As an organic combination of the Internet of Vehicles and intelligent vehicles, Intelligent Connected Vehicles (ICVs) have very high research and application value. Traditional data application methods require the local aggregation of sensitive user data, which poses a threat to user data privacy. Federated learning (FL) is a promising machine learning method that leverages distributed, personalized datasets to enhance performance while preserving user privacy. However, in mobile environments, unreliable client data can degrade the global model, reducing accuracy. Additionally, the mobility of ICVs can destabilize the training process, prolonging model updates and diminishing aggregation accuracy. To address these challenges, this paper proposes a dynamic asynchronous aggregation method that improves both reliability and training efficiency in FL for mobile networks. Therefore, it becomes crucial to find reliable aggregation of mobile device participation in FL tasks. To this end, we propose a reliable FL scheme, which only selects reliable mobile devices to participate in model aggregation to improve the generalization ability of the model. In addition, we design a dynamic asynchronous aggregation method based on reputation scores without affecting the model. Reduce model training time without compromising performance. Through experimental analysis, it is proved that this method can improve the reliability and effectiveness of FL tasks in mobile networks.

Details DOI

ICLR Conference 2025 Conference Paper

Robust Function-Calling for On-Device Language Model via Function Masking

Qiqiang Lin
Muning Wen
Qiuying Peng
Guanyu Nie
Junwei Liao
Xiaoyun Mo
Jiamu Zhou
Cheng Cheng

Large language models have demonstrated impressive value in performing as autonomous agents when equipped with external tools and API calls. Nonetheless, effectively harnessing their potential for executing complex tasks crucially relies on enhancements in their function-calling capabilities. This paper identifies a critical gap in existing function-calling models, where performance varies significantly across benchmarks, often due to over-fitting to specific naming conventions. To address such an issue, we introduce Hammer, a novel family of foundation models specifically engineered for on-device function calling. Hammer employs an augmented dataset that enhances models’ sensitivity to irrelevant functions and incorporates function masking techniques to minimize over-fitting. Our empirical evaluations reveal that Hammer not only outperforms larger models but also demonstrates robust generalization across diverse benchmarks, achieving state-of-the-art results. Our open-source contributions include a specialized dataset for irrelevance detection, a tuning framework for enhanced generalization, and the Hammer models, establishing a new standard for function-calling performance.

Details

NeurIPS Conference 2024 Conference Paper

A robust inlier identification algorithm for point cloud registration via $\mathbf{\ell_0}$-minimization

Yinuo Jiang
Xiuchuan Tang
Cheng Cheng
Ye Yuan

Correspondences in point cloud registration are prone to outliers, significantly reducing registration accuracy and highlighting the need for precise inlier identification. In this paper, we propose a robust inlier identification algorithm for point cloud registration by reformulating the conventional registration problem as an alignment error $\ell_0$-minimization problem. The $\ell_0$-minimization problem is formulated for each local set, where those local sets are built on a compatibility graph of input correspondences. To resolve the $\ell_0$-minimization, we develop a novel two-stage decoupling strategy, which first decouples the alignment error into a rotation fitting error and a translation fitting error. Second, null-space matrices are employed to decouple inlier identification from the estimation of rotation and translation respectively, thereby applying Bayesian theory to $\ell_0$-minimization problems and solving for fitting errors. Correspondences with the smallest errors are identified as inliers to generate a transformation hypothesis for each local set. The best hypothesis is selected to perform registration. We demonstrate that the proposed inlier identification algorithm is robust under high outlier ratios and noise through experiments. Extensive results on the KITTI, 3DMatch, and 3DLoMatch datasets demonstrate that our method achieves state-of-the-art performance compared to both traditional and learning-based methods in various indoor and outdoor scenes.

PDF Details DOI

NeurIPS Conference 2024 Conference Paper

Harmonizing Stochasticity and Determinism: Scene-responsive Diverse Human Motion Prediction

Tuo Wang
Zhenbo Song
Zhenyu Lou
Qiongjie Cui
Luoming Zhang
Cheng Cheng
Haofan Wang
Xu Tang

Diverse human motion prediction (HMP) is a fundamental application in computer vision that has recently attracted considerable interest. Prior methods primarily focus on the stochastic nature of human motion, while neglecting the specific impact of external environment, leading to the pronounced artifacts in prediction when applied to real-world scenarios. To fill this gap, this work introduces a novel task: predicting diverse human motion within real-world 3D scenes. In contrast to prior works, it requires harmonizing the deterministic constraints imposed by the surrounding 3D scenes with the stochastic aspect of human motion. For this purpose, we propose DiMoP3D, a diverse motion prediction framework with 3D scene awareness, which leverages the 3D point cloud and observed sequence to generate diverse and high-fidelity predictions. DiMoP3D is able to comprehend the 3D scene, and determines the probable target objects and their desired interactive pose based on the historical motion. Then, it plans the obstacle-free trajectory towards these interested objects, and generates diverse and physically-consistent future motions. On top of that, DiMoP3D identifies deterministic factors in the scene and integrates them into the stochastic modeling, making the diverse HMP in realistic scenes become a controllable stochastic generation process. On two real-captured benchmarks, DiMoP3D has demonstrated significant improvements over state-of-the-art methods, showcasing its effectiveness in generating diverse and physically-consistent motion predictions within real-world 3D environments.

PDF Details DOI

IJCAI Conference 2023 Conference Paper

Graph Propagation Transformer for Graph Representation Learning

Zhe Chen
Hao Tan
Tao Wang
Tianrun Shen
Tong Lu
Qiuying Peng
Cheng Cheng
Yue Qi

This paper presents a novel transformer architecture for graph representation learning. The core insight of our method is to fully consider the information propagation among nodes and edges in a graph when building the attention module in the transformer blocks. Specifically, we propose a new attention mechanism called Graph Propagation Attention (GPA). It explicitly passes the information among nodes and edges in three ways, i. e. node-to-node, node-to-edge, and edge-to-node, which is essential for learning graph-structured data. On this basis, we design an effective transformer architecture named Graph Propagation Transformer (GPTrans) to further help learn graph data. We verify the performance of GPTrans in a wide range of graph learning experiments on several benchmark datasets. These results show that our method outperforms many state-of-the-art transformer-based graph models with better performance. The code will be released at https: //github. com/czczup/GPTrans.

PDF Details DOI

NeurIPS Conference 2023 Conference Paper

Meta-Adapter: An Online Few-shot Learner for Vision-Language Model

Cheng Cheng
Lin Song
Ruoyi Xue
Hang Wang
Hongbin Sun
Yixiao Ge
Ying Shan

The contrastive vision-language pre-training, known as CLIP, demonstrates remarkable potential in perceiving open-world visual concepts, enabling effective zero-shot image recognition. Nevertheless, few-shot learning methods based on CLIP typically require offline fine-tuning of the parameters on few-shot samples, resulting in longer inference time and the risk of overfitting in certain domains. To tackle these challenges, we propose the Meta-Adapter, a lightweight residual-style adapter, to refine the CLIP features guided by the few-shot samples in an online manner. With a few training samples, our method can enable effective few-shot learning capabilities and generalize to unseen data or tasks without additional fine-tuning, achieving competitive performance and high efficiency. Without bells and whistles, our approach outperforms the state-of-the-art online few-shot learning method by an average of 3. 6\% on eight image classification datasets with higher inference speed. Furthermore, our model is simple and flexible, serving as a plug-and-play module directly applicable to downstream tasks. Without further fine-tuning, Meta-Adapter obtains notable performance improvements in open-vocabulary object detection and segmentation tasks.

PDF Details

JBHI Journal 2023 Journal Article

Multi-Domain Encoding of Spatiotemporal Dynamics in EEG for Emotion Recognition

Cheng Cheng
Yong Zhang
Luyao Liu
Wenzhe Liu
Lin Feng

The common goal of the studies is to map any emotional states encoded from electroencephalogram (EEG) into 2-dimensional arousal-valance scores. It is still challenging due to each emotion having its specific spatial structure and dynamic dependence over the distinct time segments among EEG signals. This paper aims to model human dynamic emotional behavior by considering the location connectivity and context dependency of brain electrodes. Thus, we designed a hybrid EEG modeling method that mainly adopts the attention mechanism, combining a multi-domain spatial transformer (MST) module and a dynamic temporal transformer (DTT) module, named MSDTTs. Specifically, the MST module extracts single-domain and cross-domain features from different brain regions and fuses them into multi-domain spatial features. Meanwhile, the temporal dynamic excitation (TDE) is inserted into the multi-head convolutional transformer to form the DTT module. These two blocks work together to activate and extract the emotion-related dynamic temporal features within the DTT module. Furthermore, we place the convolutional mapping into the transformer structure to mine the static context features among the keyframes. Overall results show that high classification accuracy of 98. 91%/0. 14% was obtained by the $\beta$ frequency band of the DEAP dataset, and 97. 52%/0. 12% and 96. 70%/0. 26% were obtained by the $\gamma$ frequency band of SEED and SEED-IV datasets. Empirical experiments indicate that our proposed method can achieve remarkable results in comparison with state-of-the-art algorithms.

Details DOI

EAAI Journal 2023 Journal Article

Randomized block-coordinate adaptive algorithms for nonconvex optimization problems

Yangfan Zhou
Kaizhu Huang
Jiang Li
Cheng Cheng
Xuguang Wang
Amir Hussian
Xin Liu

Nonconvex optimization problems have always been one focus in deep learning, in which many fast adaptive algorithms based on momentum are applied. However, the full gradient computation of high-dimensional feature vector in the above tasks become prohibitive. To reduce the computation cost for optimizers on nonconvex optimization problems typically seen in deep learning, this work proposes a randomized block-coordinate adaptive optimization algorithm, named RAda, which randomly picks a block from the full coordinates of the parameter vector and then sparsely computes its gradient. We prove that RAda converges to a δ -accurate solution with the stochastic first-order complexity of O ( 1 / δ 2 ), where δ is the upper bound of the gradient’s square, under nonconvex cases. Experiments on public datasets including CIFAR-10, CIFAR-100, and Penn TreeBank, verify that RAda outperforms the other compared algorithms in terms of the computational cost.

Details DOI

JBHI Journal 2022 Journal Article

EEG-Based Emotion Recognition Using Spatial-Temporal Graph Convolutional LSTM With Attention Mechanism

Lin Feng
Cheng Cheng
Mingyan Zhao
Huiyuan Deng
Yong Zhang

The dynamic uncertain relationship among each brain region is a necessary factor that limits EEG-based emotion recognition. It is a thought-provoking problem to availably employ time-varying spatial and temporal characteristics from multi-channel electroencephalogram (EEG) signals. Although deep learning has made remarkable achievements in emotion recognition, the biological topological information among brain regions does not fully exploit, which is vital for EEG-based emotion recognition. In response to this problem, we design a hybrid model called ST-GCLSTM, which comprises a spatial-graph convolutional network (SGCN) module and an attention-enhanced bi-directional Long Short-Term Memory (LSTM) module. The main advantage of ST-GCLSTM is that it can consider the biological topology information of each brain region to extract representative spatial-temporal features from multiple EEG channels. Specifically, we construct two layers SGCN by introducing adjacency matrices to adaptively learn the intrinsic connection among different EEG channels. Moreover, an attention-enhanced mechanism is placed into a bi-directional LSTM module to extract the crucial spatial-temporal features from sequential EEG data, and then these features serve as the input layer of the classifier to learn discriminative emotion-related features. Extensive experiments on the DEAP, SEED, and SEED-IV datasets demonstrate the effectiveness of the proposed ST-GCLSTM model, revealing that our model had an absolute performance improvement over state-of-the-art strategies.

Details DOI

AAAI Conference 2020 System Paper

Combining Machine Learning Models Using combo Library

Yue Zhao
Xuejian Wang
Cheng Cheng
Xueying Ding

Model combination, often regarded as a key sub-ﬁeld of ensemble learning, has been widely used in both academic research and industry applications. To facilitate this process, we propose and implement an easy-to-use Python toolkit, combo, to aggregate models and scores under various scenarios, including classiﬁcation, clustering, and anomaly detection. In a nutshell, combo provides a uni- ﬁed and consistent way to combine both raw and pretrained models from popular machine learning libraries, e. g. , scikit-learn, XGBoost, and LightGBM. With accessibility and robustness in mind, combo is designed with detailed documentation, interactive examples, continuous integration, code coverage, and maintainability check; it can be installed easily through Python Package Index (PyPI) or https: //github. com/yzhao062/combo.

PDF Details

IS Journal 2010 Journal Article

Parallel Traffic Management for the 2010 Asian Games

Gang Xiong
Kunfeng Wang
Fenghua Zhu
Cheng Cheng
Xiangjing An
Zhendong Xie

The 16th Asian Games will be held in Guangzhou, China, in November 2010. With 42 sports scheduled, more than 100 thousand people directly involved, and a scale similar to Beijing's 2008 Olympic Games, it will be the largest Asian Games event. One of the major challenges facing the 2010 Asian Games is the transportation problem. Because the games will use 58 existing game facilities and 12 new sports stadiums, which are located across the Guangzhou metropolitan areas, safe and effective traffic control and transportation management will be essential to their success. Under such time and infrastructure constraints, Guangzhou has chosen parallel transportation management systems to enhance and improve its traffic safety and efficiency for a better 2010 Asian Games,

Details DOI