Xu Cheng

AAAI Conference 2026 Conference Paper

Beyond Missing Data Imputation: Information-Theoretic Coupling of Missingness and Class Imbalance for Optimal Irregular Time Series Classification

• Xin Qin
• Mengna Liu
• Wenjie Wang
• Shuxin Li
• Tianjiao Li
• Xiufeng Liu
• Xu Cheng

Irregular time series (IRTS) are prevalent in real-world applications, where uneven sampling and missing data pose fundamental challenges to deep learning-based feature modeling. Although existing methods attempt to retain timestamp information, they often overlook the structured patterns embedded within the missingness itself, and tend to perform poorly when confronted with class imbalance exacerbated by data incompleteness. Specifically, temporal irregularity hinders the modeling of long-range dependencies and local patterns, while sparse observations limit representational capacity, disproportionately impairing minority classes and leading to severe classification bias. To address these deeply coupled challenges, we propose SPECTRA (Structured Pattern and Enriched Context-aware Temporal Representation Architecture), a unified framework for robust IRTS classification. SPECTRA introduces a frequency-guided observation encoder that reconstructs temporal dependencies in a stable manner, mitigating spectral distortion and information corruption. Complementarily, a missingness pattern encoder explicitly captures the dynamic evolution of missing data and leverages it as a discriminative signal. In addition, a prototype-constrained classification paradigm directly optimizes the geometric structure of the feature space, enhancing intra-class compactness and alleviating generalization bottlenecks caused by class imbalance. Extensive experiments on three public IRTS datasets—P12, P19, and PAM—demonstrate the superior performance of SPECTRA under both missing and imbalanced conditions.

EAAI Journal 2026 Journal Article

Corrigendum to “Physics-informed dynamic ensemble learning for real-time urban water quality monitoring” [Eng. Appl. Artif. Intell. 175 (2026) 114628]

• Renfang Wang
• Xinyu Zhao
• Xiufeng Liu
• Xu Cheng
• Hong Qiu

EAAI Journal 2026 Journal Article

Learning invariant representation for light field adversarial salient object detection

• Mianzhao Wang
• Fan Shi
• Xu Cheng
• Shengyong Chen

Pixel-wise salient object detection (SOD) in light fields is highly vulnerable to adversarial attacks, as attacking pixels across multiple focal planes makes it difficult to preserve consistent salient regions. To address this challenge, we propose a novel Artificial Intelligence (AI) framework, Masked Light Field Encoding (MLFE), which integrates a mask generator and a Cross-Focal Slices Exchange (CFSE) mechanism to learn invariant light field representations during feature reconstruction. The mask generator suppresses most adversarially perturbed pixels, forcing the encoder to recover the underlying semantic distribution from limited features. Meanwhile, CFSE exchanges feature patches across focal slices, leveraging dynamic cues to capture contextual consistency. Optimized with clean samples, MLFE acquires robust and invariant light field features, significantly improving resistance to adversarial perturbations. Empirical evaluations demonstrate that our method not only surpasses state-of-the-art approaches under adversarial conditions but also enhances robustness on clean data, highlighting the effectiveness of AI-driven light field SOD in real-world applications.

EAAI Journal 2026 Journal Article

Physics-informed dynamic ensemble learning for real-time urban water quality monitoring

• Renfang Wang
• Xinyu Zhao
• Xiufeng Liu
• Xu Cheng
• Hong Qiu

Ensuring high-quality water resources is crucial for sustainable urban development, public health, and resilient city infrastructure, yet traditional anomaly detection methods struggle with the highly variable, non-stationary, and concept-drifting nature of urban water quality data streams. This study proposes a Physics-Informed Dynamic Ensemble Learning (PIDEL) framework, an artificial intelligence approach that combines diverse classical and deep learning models with Physics-Informed Neural Networks (PINNs) embedding convection–diffusion constraints, a Genetic Algorithm (GA) for ensemble optimization, and a Jensen–Shannon Divergence (JSD) based mechanism for dynamic model switching. Applied to a real-world urban water quality dataset, PIDEL achieves an F1-score of 0. 95, representing a 59% improvement over the best static ensemble, while reducing false alarms by 73% compared to traditional methods and maintaining F1-scores above 0. 9 across all sliding windows. The framework processes each 60-minute window in approximately 2. 3 s on standard hardware, demonstrating its suitability for real-time deployment in smart city water systems. These results highlight that integrating physics-informed constraints with dynamic ensemble learning can substantially enhance the reliability, interpretability, and operational value of automated water quality anomaly detection for urban utilities.

AAAI Conference 2026 Conference Paper

Radar-APLANC: Unsupervised Radar-based Heartbeat Sensing via Augmented Pseudo-Label and Noise Contrast

• Ying Wang
• Zhaodong Sun
• Xu Cheng
• Zuxian He
• Xiaobai Li

Frequency Modulated Continuous Wave (FMCW) radars can measure subtle chest wall oscillations to enable non-contact heartbeat sensing. However, traditional radar-based heartbeat sensing methods face performance degradation due to noise. Learning-based radar methods achieve better noise robustness but require costly labeled signals for supervised training. To overcome these limitations, we propose the first unsupervised framework for radar-based heartbeat sensing via Augmented Pseudo-Label and Noise Contrast (Radar-APLANC). We propose to use both the heartbeat range and noise range within the radar range matrix to construct the positive and negative samples, respectively, for improved noise robustness. Our Noise-Contrastive Triplet (NCT) loss only utilizes positive samples, negative samples, and pseudo-label signals generated by the traditional radar method, thereby avoiding dependence on expensive ground-truth physiological signals. We further design a pseudo-label augmentation approach featuring adaptive noise-aware label selection to improve pseudo-label signal quality. Extensive experiments on the Equipleth dataset and our collected radar dataset demonstrate that our unsupervised method achieves performance comparable to state-of-the-art supervised methods.

EAAI Journal 2025 Journal Article

An expert features enhanced temporal and contextual contrasting learning model for detecting wind turbine blade icing

• Jiamei Zhou
• Feng Xiao
• Xiaoying Zhang
• Xu Cheng
• Jianhua Zhang

With global carbon neutrality goals, wind power has rapidly developed, but blade icing remains a major challenge. AI(artificial intelligence) methods show great promise for detecting icing on wind turbine blades. However, early icing data overlap, difficulty obtaining continuous labeled data, and small variations between samples due to short sampling intervals complicate the task. This study proposes an expert feature-enhanced temporal and contextual contrastive learning model for detecting blade icing. This approach efficiently extracts data features and combines self-supervised contrastive learning, maximizing data utilization without requiring extensive labeled data. To validate the effectiveness of this method, extensive experiments were conducted on two public datasets. The results achieved the best performance across multiple metrics, with F1-Score and AUC exceeding 98%, significantly enhancing wind power generation efficiency.

AAAI Conference 2025 Conference Paper

Can Students Beyond the Teacher? Distilling Knowledge from Teacher’s Bias

• Jianhua Zhang
• Yi Gao
• Ruyu Liu
• Xu Cheng
• Houxiang Zhang
• Shengyong Chen

Knowledge distillation (KD) is a model compression technique that transfers knowledge from a large teacher model to a smaller student model to enhance its performance. Existing methods often assume that the student model is inherently inferior to the teacher model. However, we identify that the fundamental issue affecting student performance is the bias transferred by the teacher. Current KD frameworks transmit both right and wrong knowledge, introducing bias that misleads the student model. To address this issue, we propose a novel strategy to rectify bias and greatly improve the student model's performance. Our strategy involves three steps: First, we differentiate knowledge and design a bias elimination method to filter out biases, retaining only the right knowledge for the student model to learn. Next, we propose a bias rectification method to rectify the teacher model's wrong predictions, fundamentally addressing bias interference. The student model learns from both the right knowledge and the rectified biases, greatly improving its prediction accuracy. Additionally, we introduce a dynamic learning approach with a loss function that updates weights dynamically, allowing the student model to quickly learn right knowledge-based easy tasks initially and tackle hard tasks corresponding to biases later, greatly enhancing the student model's learning efficiency. To the best of our knowledge, this is the first strategy enabling the student model to surpass the teacher model. Experiments demonstrate that our strategy, as a plug-and-play module, is versatile across various mainstream KD frameworks.

EAAI Journal 2025 Journal Article

Enhancing spatiotemporal wind power forecasting with meta-learning in data-scarce environments

• Renfang Wang
• Jingtong Wu
• Xu Cheng
• Xiufeng Liu
• Hong Qiu

Accurate wind power forecasting is critical for maintaining stable power grids, yet the inherent variability of wind and limited data availability for new wind farms present significant challenges. To address these issues, we present a novel artificial intelligence framework that integrates a self-attention enhanced Spatiotemporal Long Short-Term Memory (ST-LSTM) network with Model-Agnostic Meta-Learning (MAML), termed as the Meta-Learning Spatiotemporal Attention Long Short-Term Memory framework (MAML-STALSTM). This deep learning combination enables the model to effectively capture long-range spatiotemporal dependencies while rapidly adapting to new wind farm configurations or changing wind conditions with minimal training data. By employing rigorous data preprocessing techniques and ensuring temporal separation in data splitting, we mitigate potential data leakage and enhance the model’s generalizability. Extensive experiments conducted on both onshore and offshore wind farm datasets demonstrate that our artificial intelligence approach outperforms established baseline models, particularly excelling in data-scarce environments. Ablation studies highlight the crucial roles of the self-attention mechanism and meta-learning in improving forecasting accuracy, adaptation speed, and model robustness. These results emphasize the practical benefits of our approach in enhancing grid stability and supporting the seamless integration of wind energy, thereby contributing significantly to the advancement of sustainable energy solutions.

AAAI Conference 2025 Conference Paper

FreeNet: Liberating Depth-Wise Separable Operations for Building Faster Mobile Vision Architectures

• Hao Yu
• Haoyu Chen
• Wei Peng
• Xu Cheng
• Guoying Zhao

In the pursuit of efficient vision architectures, substantial efforts have been devoted to optimizing operator efficiency. Depth-wise separable operators, such as DWConv, are found cheap in both FLOPs and parameters. As a result, they are increasingly incorporated into efficient backbones, trading for deeper and wider architectures to enhance performance. However, separable operators are not really fast on devices due to the discontinuous memory access requirements. In this paper, we propose FreeNets, a family of simple and efficient backbones that free the separable operation to further accelerate the running speed. We introduce sparse sampling mixers (S2-Mixer) to supersede existing separable token mixers. The S2-Mixer samples multiple segments of partially continuous signals across spatial and channel dimensions for convolutional processing, achieving extremely fast on-device speed. The sparse sampling also enables S2-Mixer to capture long-range pixel relationships from dynamic receptive fields. Furthermore, we introduce a Shift Feed-Forward Network (ShiftFFN) as a faster alternative to existing channel mixers. It utilizes a shift neck architecture that aggregates global information to shift features, enabling faster channel mixing while incorporating global pixel information. Extensive experiments demonstrate that FreeNet offers a superior accuracy-efficiency tradeoff compared to the latest efficient models. On ImageNet-1k, FreeNet-S2 outperforms the StarNet-S4 by 0.4% in top-1 accuracy, while running around 40% faster on desktop GPU and 15% faster on Mobile GPU.

NeurIPS Conference 2025 Conference Paper

Multi-order Orchestrated Curriculum Distillation for Model-Heterogeneous Federated Graph Learning

• Frank Wan
• Xu Cheng
• Run Liu
• Wenke Huang
• Zitong Shi
• Pinyi Jin
• Guibin Zhang
• Bo Du

Federated Graph Learning (FGL) has been shown to be particularly effective in enabling collaborative training of Graph Neural Networks (GNNs) in decentralized settings. Model-heterogeneous FGL further enhances practical applicability by accommodating client preferences for diverse model architectures. However, existing model-heterogeneous approaches primarily target Euclidean data and fail to account for a crucial aspect of graph-structured data: topological relationships. To address this limitation, we propose **TRUST**, a novel knowledge distillation-based **model-heterogeneous FGL** framework. Specifically, we propose Progressive Curriculum Node Scheduler to progressively introduce challenging nodes based on learning difficulty. In Adaptive Curriculum Distillation Modulator, we propose an adaptive temperature modulator that dynamically adjusts knowledge distillation temperature to accommodate varying client capabilities and graph complexity. Moreover, we leverage Wasserstein‑Driven Affinity Distillation to enable models to capture cross-class structural relationships through optimal transport. Extensive experiments on multiple graph benchmarks and model-heterogeneous settings show that **TRUST** outperforms existing methods, achieving an average 3. 6\% $\uparrow$ performance gain, particularly under moderate heterogeneity conditions. The code is available for anonymous access at https: //anonymous. 4open. science/r/TRUST-NeurIPS2025.

JBHI Journal 2025 Journal Article

Non-invasive fECG monitoring through multifaceted temporal feature interaction of maternal abdominal electrocardiography

• Yiwei Lin
• Zhenqin Chen
• Xu Cheng
• Qiong Luo
• Jinshan Xu

Fetal electrocardiography (fECG) signals are an essential non-invasive indicator for monitoring fetal cardiac health. However, extracting high-quality fECG signals from maternal abdominal electrocardiography (aECG) remains challenging due to severe interference from maternal electrocardiography (mECG) and ambient noise. Existing fECG extracting methods have achieved remarkable performance. However, most works fall short in modeling the long-term dynamic changes between consecutive heartbeats, resulting in low discrimination between different heartbeats with similar semantic information, such as rhythm and amplitude variations. To this end, we propose a dual-branch architecture that accurately extracts fECG signals by combining the efficiency of CNNs in extracting local heartbeat information with the capability of Transformers to model long-term dependencies between consecutive heartbeats. A collaborative feature interaction unit is further introduced to enhance information communication within the parallel branches and exploit the complementarity between detailed morphological patterns and long-range temporal dependencies. Experimental results on two public datasets showcase the exceptional performance of our proposed method, achieving Pearson correlation coefficient of 0. 90 on ADDB and 0. 94 on BDDB-L in assessing fECG signal quality. These findings suggest that our approach offers a promising solution for continuous fetal heart monitoring.

NeurIPS Conference 2025 Conference Paper

PolypSense3D: A Multi-Source Benchmark Dataset for Depth-Aware Polyp Size Measurement in Endoscopy

• Ruyu Liu
• Lin Wang
• Zhou Mingming
• Jianhua Zhang
• ZHANG HAOYU
• Xiufeng Liu
• Xu Cheng
• Sixian Chan

Accurate polyp sizing during endoscopy is crucial for cancer risk assessment but is hindered by subjective methods and inadequate datasets lacking integrated 2D appearance, 3D structure, and real-world size information. We introduce PolypSense3D, the first multi-source benchmark dataset specifically targeting depth-aware polyp size measurement. It uniquely integrates over 43, 000 frames from virtual simulations, physical phantoms, and clinical sequences, providing synchronized RGB, dense/sparse depth, segmentation masks, camera parameters, and millimeter-scale size labels derived via a novel forceps-assisted in-vivo annotation technique. To establish its value, we benchmark state-of-the-art segmentation and depth estimation models. Results quantify significant domain gaps between simulated/phantom and clinical data and reveal substantial error propagation from perception stages to final size estimation, with the best fully automated pipelines achieving an average Mean Absolute Error (MAE) of 0. 95 mm on the clinical data subset. Publicly released under CC BY-SA 4. 0 with code and evaluation protocols, PolypSense3D offers a standardized platform to accelerate research in robust, clinically relevant quantitative endoscopic vision. The benchmark dataset and code are available at: https: //github. com/HNUicda/PolypSense3D and https: //doi. org/10. 7910/DVN/K13H89.

AAAI Conference 2024 Conference Paper

Clarifying the Behavior and the Difficulty of Adversarial Training

• Xu Cheng
• Hao Zhang
• Yue Xin
• Wen Shen
• Quanshi Zhang

Adversarial training is usually difficult to optimize. This paper provides conceptual and analytic insights into the difficulty of adversarial training via a simple theoretical study, where we derive an approximate dynamics of a recursive multi-step attack in a simple setting. Despite the simplicity of our theory, it still reveals verifiable predictions about various phenomena in adversarial training under real-world settings. First, compared to vanilla training, adversarial training is more likely to boost the influence of input samples with large gradient norms in an exponential manner. Besides, adversarial training also strengthens the influence of the Hessian matrix of the loss w.r.t. network parameters, which is more likely to make network parameters oscillate and boosts the difficulty of adversarial training.

AAAI Conference 2024 Conference Paper

Differentiable Auxiliary Learning for Sketch Re-Identification

• Xingyu Liu
• Xu Cheng
• Haoyu Chen
• Hao Yu
• Guoying Zhao

Sketch re-identification (Re-ID) seeks to match pedestrians' photos from surveillance videos with corresponding sketches. However, we observe that existing works still have two critical limitations: (i) cross- and intra-modality discrepancies hinder the extraction of modality-shared features, (ii) standard triplet loss fails to constrain latent feature distribution in each modality with inadequate samples. To overcome the above issues, we propose a differentiable auxiliary learning network (DALNet) to explore a robust auxiliary modality for Sketch Re-ID. Specifically, for (i) we construct an auxiliary modality by using a dynamic auxiliary generator (DAG) to bridge the gap between sketch and photo modalities. The auxiliary modality highlights the described person in photos to mitigate background clutter and learns sketch style through style refinement. Moreover, a modality interactive attention module (MIA) is presented to align the features and learn the invariant patterns of two modalities by auxiliary modality. To address (ii), we propose a multi-modality collaborative learning scheme (MMCL) to align the latent distribution of three modalities. An intra-modality circle loss in MMCL brings learned global and modality-shared features of the same identity closer in the case of insufficient samples within each modality. Extensive experiments verify the superior performance of our DALNet over the state-of-the-art methods for Sketch Re-ID, and the generalization in sketch-based image retrieval and sketch-photo face recognition tasks.

JBHI Journal 2024 Journal Article

ST-Phys: Unsupervised Spatio-Temporal Contrastive Remote Physiological Measurement

• Mingyue Cao
• Xu Cheng
• Xingyu Liu
• Yan Jiang
• Hao Yu
• Jingang Shi

Remote photoplethysmography (rPPG) is a non-contact method that employs facial videos for measuring physiological parameters. Existing rPPG methods have achieved remarkable performance. However, the success mainly profits from supervised learning over massive labeled data. On the other hand, existing unsupervised rPPG methods fail to fully utilize spatio-temporal features and encounter challenges in low-light or noise environments. To address these problems, we propose an unsupervised contrast learning approach, ST-Phys. We incorporate a low-light enhancement module, a temporal dilated module, and a spatial enhanced module to better deal with long-term dependencies under the random low-light conditions. In addition, we design a circular margin loss, wherein rPPG signals originating from identical videos are attracted, while those from distinct videos are repelled. Our method is assessed on six openly accessible datasets, including RGB and NIR videos. Extensive experiments reveal the superior performance of our proposed ST-Phys over state-of-the-art unsupervised rPPG methods. Moreover, it offers advantages in parameter reduction and noise robustness.

NeurIPS Conference 2023 Conference Paper

Towards the Difficulty for a Deep Neural Network to Learn Concepts of Different Complexities

• Dongrui Liu
• Huiqi Deng
• Xu Cheng
• Qihan Ren
• Kangrui Wang
• Quanshi Zhang

This paper theoretically explains the intuition that simple concepts are more likely to be learned by deep neural networks (DNNs) than complex concepts. In fact, recent studies have observed [24, 15] and proved [26] the emergence of interactive concepts in a DNN, i. e. , it is proven that a DNN usually only encodes a small number of interactive concepts, and can be considered to use their interaction effects to compute inference scores. Each interactive concept is encoded by the DNN to represent the collaboration between a set of input variables. Therefore, in this study, we aim to theoretically explain that interactive concepts involving more input variables (i. e. , more complex concepts) are more difficult to learn. Our finding clarifies the exact conceptual complexity that boosts the learning difficulty.

YNICL Journal 2022 Journal Article

Changes in sensory-related brain networks of patients with moyamoya disease with limb paresthesia: A resting-state fMRI-based functional connectivity analysis

• Rujing Sun
• Shi-Yu Zhang
• Xu Cheng
• Peng Zhang
• Peng-Gang Qiao
• Gong-Jie Li

This study's aim was to investigate functional brain connectivity changes among patients with moyamoya disease (MMD) with limb paresthesia, using functional connectivity analysis based on resting-state functional magnetic resonance imaging (rs-fMRI). A total of 181 patients with MMD were enrolled, including 57 with left limb paresthesia (MLP group), 61 with right limb paresthesia (MRP group), and 63 without paresthesia (MWP group). Encephaloduroarteriosynangiosis (EDAS) was performed in 20 of the 57 patients with left limb paresthesia and 15 of the 61 patients with right limb paresthesia. Twenty-nine age- and sex-matched healthy controls (HC group) were recruited during the same period. All participants underwent rs-fMRI examination, and the patients treated with EDAS were re-examined 3-4 months after the surgery. After data preprocessing, we selected Brodmann area 3 on each side of the brain as the seed region to construct a functional connectivity network of the whole brain, and then we analyzed the differences in functional connectivity between the HC group, MWP group, MLP group, and MRP group. The functional connectivity of Brodmann area 3 (on either side) with the ipsilateral frontal (superior frontal gyrus, middle frontal gyrus, and inferior frontal gyrus) and parietal (supramarginal gyrus, angular gyrus, and superior parietal lobule) cortices was increased among patients with MMD. The functional connectivity enhancement in these brain regions was broader and greater in patients with contralateral limb paresthesia than in patients without paresthesia, and the regions with functional connectivity changes were roughly distributed symmetrically among the MLP group and the MRP group. There were no changes 3-4 months after EDAS in the increased functional connectivity between the frontal and parietal cortices and Brodmann area 3. Limb paresthesia in patients with MMD may be driven by abnormal functional connectivity in the frontal and parietal cortices. Functional changes in associated brain regions may be a target for evaluating the severity of MMD and its response to treatment.

AAAI Conference 2021 Conference Paper

Building Interpretable Interaction Trees for Deep NLP Models

• Die Zhang
• Hao Zhang
• Huilin Zhou
• Xiaoyi Bao
• Da Huo
• Ruizhao Chen
• Xu Cheng
• Mengyue Wu

This paper proposes a method to disentangle and quantify interactions among words that are encoded inside a DNN for natural language processing. We construct a tree to encode salient interactions extracted by the DNN. Six metrics are proposed to analyze properties of interactions between constituents in a sentence. The interaction is defined based on Shapley values of words, which are considered as an unbiased estimation of word contributions to the network prediction. Our method is used to quantify word interactions encoded inside the BERT, ELMo, LSTM, CNN, and Transformer networks. Experimental results have provided a new perspective to understand these DNNs, and have demonstrated the effectiveness of our method.

IJCAI Conference 2021 Conference Paper

Drop Redundant, Shrink Irrelevant: Selective Knowledge Injection for Language Pretraining

• Ningyu Zhang
• Shumin Deng
• Xu Cheng
• Xi Chen
• Yichi Zhang
• Wei Zhang
• Huajun Chen

Previous research has demonstrated the power of leveraging prior knowledge to improve the performance of deep models in natural language processing. However, traditional methods neglect the fact that redundant and irrelevant knowledge exists in external knowledge bases. In this study, we launched an in-depth empirical investigation into downstream tasks and found that knowledge-enhanced approaches do not always exhibit satisfactory improvements. To this end, we investigate the fundamental reasons for ineffective knowledge infusion and present selective injection for language pretraining, which constitutes a model-agnostic method and is readily pluggable into previous approaches. Experimental results on benchmark datasets demonstrate that our approach can enhance state-of-the-art knowledge injection methods.

NeurIPS Conference 2021 Conference Paper

Towards a Unified Game-Theoretic View of Adversarial Perturbations and Robustness

• Jie Ren
• Die Zhang
• Yisen Wang
• Lu Chen
• Zhanpeng Zhou
• Yiting Chen
• Xu Cheng
• Xin Wang

This paper provides a unified view to explain different adversarial attacks and defense methods, i. e. the view of multi-order interactions between input variables of DNNs. Based on the multi-order interaction, we discover that adversarial attacks mainly affect high-order interactions to fool the DNN. Furthermore, we find that the robustness of adversarially trained DNNs comes from category-specific low-order interactions. Our findings provide a potential method to unify adversarial perturbations and robustness, which can explain the existing robustness-boosting methods in a principle way. Besides, our findings also make a revision of previous inaccurate understanding of the shape bias of adversarially learned features. Our code is available online at https: //github. com/Jie-Ren/A-Unified-Game-Theoretic-Interpretation-of-Adversarial-Robustness.