Hongying Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Hongying Zhang has a total of 52 co-authors. Here are the top co-authors by number of shared publications:
- Fei Hao
Shaanxi Normal University
4 shared publications - Kezhen Dong
Xi’an Jiaotong University
4 shared publications - Syed Masroor Anwar
Department of Statistics University of Azad Jammu and Kashmir Muzaffarabad Pakistan
4 shared publications - Zahid Rasheed
Department of Mathematics, Women University of Azad Jammu and Kashmir Bagh, Bagh, Pakistan
4 shared publications - Deyu Meng
Xi’an Jiaotong University
3 shared publications - Jiang Xue
School of Mathematics and Statistics, Xian Jiaotong University, Xian, China
3 shared publications
Latest Publications
- 8 August 2026
- Published by Elsevier in Information Fusion
- Vol. 138
TARGET: Efficient and generalizable multi-view partial multi-label learning via anchor graphs and high-order tensor correlation
- 8 August 2026
- Published by Elsevier in Information Fusion
- Vol. 138
TARGET: Efficient and generalizable multi-view partial multi-label learning via anchor graphs and high-order tensor correlation
- 14 March 2026
- Published by Elsevier in Pattern Recognition
- Vol. 178
FastGRAIL: Anchor-graph-based fast adaptive partial multi-label learning with label correlations
- 14 March 2026
- Published by Elsevier in Pattern Recognition
- Vol. 178
FastGRAIL: Anchor-graph-based fast adaptive partial multi-label learning with label correlations
- 13 March 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 319
AG2CD: Soft anchor graph based fast community detection in attributed graphs
- 13 March 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 319
AG2CD: Soft anchor graph based fast community detection in attributed graphs
- 20 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. PP, 1-13
Vertex Entanglement-Enhanced Graph Contrastive Learning for Social Networks
Graph contrastive learning can extract high-quality representations from graph-structured data, providing an effective way to analyze social networks. However, existing graph contrastive learning methods with both random augmentations and advanced adaptive strategies struggle to effectively preserve key structural semantics and eliminate graph noise, which ultimately limits representation quality. To this end, this article proposes a vertex entanglement-enhanced graph contrastive learning (VEGCL), which utilizes vertex entanglement (VE) as a novel metric that quantifies node importance from a global structural perspective to guide the removal of unimportant nodes and edges for generating augmented views. Specifically, VE is initially employed to selectively remove less important nodes, generating the first augmented view. Subsequently, the original graph is transformed into a line graph, after which VE is again applied to remove less critical edges, producing the second augmented view. Following augmentation, model training is guided by the Info Noise Contrastive Estimation loss function to enhance embedding quality. To verify the effectiveness of the proposed approach, node classification experiments are conducted on five commonly used benchmark datasets including Cora, PubMed, CiteSeer, Coauthor-CS, and LastFM Asia. The experimental results demonstrate that the proposed method improves the performance of node classification by $0.5\%\boldsymbol{\sim} 2.3\%$ , and $0.04\%\boldsymbol{\sim} 5\%$ , in terms of accuracy and F1 score compared with the existing baseline methods.
- 20 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. PP, 1-13
Vertex Entanglement-Enhanced Graph Contrastive Learning for Social Networks
Graph contrastive learning can extract high-quality representations from graph-structured data, providing an effective way to analyze social networks. However, existing graph contrastive learning methods with both random augmentations and advanced adaptive strategies struggle to effectively preserve key structural semantics and eliminate graph noise, which ultimately limits representation quality. To this end, this article proposes a vertex entanglement-enhanced graph contrastive learning (VEGCL), which utilizes vertex entanglement (VE) as a novel metric that quantifies node importance from a global structural perspective to guide the removal of unimportant nodes and edges for generating augmented views. Specifically, VE is initially employed to selectively remove less important nodes, generating the first augmented view. Subsequently, the original graph is transformed into a line graph, after which VE is again applied to remove less critical edges, producing the second augmented view. Following augmentation, model training is guided by the Info Noise Contrastive Estimation loss function to enhance embedding quality. To verify the effectiveness of the proposed approach, node classification experiments are conducted on five commonly used benchmark datasets including Cora, PubMed, CiteSeer, Coauthor-CS, and LastFM Asia. The experimental results demonstrate that the proposed method improves the performance of node classification by $0.5\%\boldsymbol{\sim} 2.3\%$ , and $0.04\%\boldsymbol{\sim} 5\%$ , in terms of accuracy and F1 score compared with the existing baseline methods.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 19 (19391404) , 14776-14794
LassoNet-GNN: A Catalyst for Few-Shot Distorted SAR Image Classification
Few-shot learning (FSL) has revolutionized SAR image classification by enabling robust recognition with limited high-quality labeled data. However, existing methods struggle with both identifying key features in noisy, high-variation SAR data and handling geometric distortions. To address these challenges, we propose a novel contrastive sparse few-shot distorted SAR image classification framework via LassoNet-GNN, which incorporates GNN-based FSL framework with an adopted episodic training protocol and a LassoNet-based sparsification technique into an end-to-end architecture for SAR image classification and interpretable feature selection with an optional plug-and-play contrastive learning (CL) regularization module that can be beneficial for improving robustness to geometric distortions on more complex data with richer intraclass variations in few-shot scenarios. Specifically, we first build upon a GNN-based FSL framework trained with an adopted episodic protocol that forms a learnable similarity graph over features extracted by a convolutional encoder, enabling effective message passing and strong base recognition performance under limited data. In addition, we introduce a LassoNet-type input-to-output skip-layer across the entire GNN-based network and apply a sparse penalty to its weights, which encourages the weights of unimportant nodes in the first hidden layer of the network to approach zero. Finally, an auxiliary plug-and-play contrastive regularization module is employed by utilizing geometric transformation augmentations to construct positive and negative sample pairs, which can be optionally activated on complex few-shot data when beneficial for robustness. Experimental results on common few-shot SAR image datasets verify the effectiveness and efficiency of the proposed framework in classification accuracy and interpretability.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 19 (19391404) , 14776-14794
Open AccessLassoNet-GNN: A Catalyst for Few-Shot Distorted SAR Image Classification
Few-shot learning (FSL) has revolutionized SAR image classification by enabling robust recognition with limited high-quality labeled data. However, existing methods struggle with both identifying key features in noisy, high-variation SAR data and handling geometric distortions. To address these challenges, we propose a novel contrastive sparse few-shot distorted SAR image classification framework via LassoNet-GNN, which incorporates GNN-based FSL framework with an adopted episodic training protocol and a LassoNet-based sparsification technique into an end-to-end architecture for SAR image classification and interpretable feature selection with an optional plug-and-play contrastive learning (CL) regularization module that can be beneficial for improving robustness to geometric distortions on more complex data with richer intraclass variations in few-shot scenarios. Specifically, we first build upon a GNN-based FSL framework trained with an adopted episodic protocol that forms a learnable similarity graph over features extracted by a convolutional encoder, enabling effective message passing and strong base recognition performance under limited data. In addition, we introduce a LassoNet-type input-to-output skip-layer across the entire GNN-based network and apply a sparse penalty to its weights, which encourages the weights of unimportant nodes in the first hidden layer of the network to approach zero. Finally, an auxiliary plug-and-play contrastive regularization module is employed by utilizing geometric transformation augmentations to construct positive and negative sample pairs, which can be optionally activated on complex few-shot data when beneficial for robustness. Experimental results on common few-shot SAR image datasets verify the effectiveness and efficiency of the proposed framework in classification accuracy and interpretability.
- 5 May 2025
- Published by Elsevier in International Journal of Approximate Reasoning
- Vol. 184
Three-way space structure and clustering of categorical data
- 5 May 2025
- Published by Elsevier in International Journal of Approximate Reasoning
- Vol. 184
Three-way space structure and clustering of categorical data
- 18 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 33 (7) , 2284-2294
Robust Jointly Sparse Fast Fuzzy Clustering via Ternary-Tree-Based Anchor Graph
Traditional partition-based fuzzy clustering algorithms are widely used for revealing possible hidden structures in data. However, high computational cost limits their applications in large-scale and high-dimensional data. Moreover, most fuzzy clustering algorithms are sensitive to noise. To tackle these issues, a robust jointly sparse fast fuzzy clustering algorithm via anchor graph (RSFCAG) is proposed and analyzed in this paper. Specifically, we first propose a fast k-means method integrated shadowed set and balanced ternary tree, which serves as a fast hierarchical clustering approach by partitioning every cluster into three subclusters at each layer (3KHK). 3KHK can quickly obtain the anchor set and its optimization is solved fast by the simplex method, which also captures the ambiguity and uncertainty between clusters in large-scale clustering tasks. Second, a similarity matrix learning approach based on possibilistic neighbors is further proposed to get a robust similarity graph, which strengthens the ability of fuzzy clustering to handle large-scale data. Furthermore, the orthogonal projection matrix is integrated into the RSFCAG framework to transform the original high-dimensional space into low-dimensional space. Finally, the $L_{2,1}$-norm loss and regularization are integrated into the joint algorithm RSFCAG, which is solved optimally by block coordinate technique, to enhance the robustness and interpretability of the fuzzy clustering process. The experimental results demonstrate the effectiveness and efficiency of our proposed method in most of benchmark datasets.
- 18 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 33 (7) , 2284-2294
Robust Jointly Sparse Fast Fuzzy Clustering via Ternary-Tree-Based Anchor Graph
Traditional partition-based fuzzy clustering algorithms are widely used for revealing possible hidden structures in data. However, high computational cost limits their applications in large-scale and high-dimensional data. Moreover, most fuzzy clustering algorithms are sensitive to noise. To tackle these issues, a robust jointly sparse fast fuzzy clustering algorithm via anchor graph (RSFCAG) is proposed and analyzed in this paper. Specifically, we first propose a fast k-means method integrated shadowed set and balanced ternary tree, which serves as a fast hierarchical clustering approach by partitioning every cluster into three subclusters at each layer (3KHK). 3KHK can quickly obtain the anchor set and its optimization is solved fast by the simplex method, which also captures the ambiguity and uncertainty between clusters in large-scale clustering tasks. Second, a similarity matrix learning approach based on possibilistic neighbors is further proposed to get a robust similarity graph, which strengthens the ability of fuzzy clustering to handle large-scale data. Furthermore, the orthogonal projection matrix is integrated into the RSFCAG framework to transform the original high-dimensional space into low-dimensional space. Finally, the $L_{2,1}$-norm loss and regularization are integrated into the joint algorithm RSFCAG, which is solved optimally by block coordinate technique, to enhance the robustness and interpretability of the fuzzy clustering process. The experimental results demonstrate the effectiveness and efficiency of our proposed method in most of benchmark datasets.
- 27 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 24 (8) , 6413-6428
Variable-Depth Learning Architecture to Adaptive Multi-User Channel Prediction
On the road to 6G, the growing number of antennas and mobility-related applications emphasize the urgency of addressing the channel aging issue. Traditional channel prediction methods are no longer valid due to simplified assumptions. Deep learning (DL)-based predictors, even with attractive performance improvements, generally lack the utilization of extra correlations and flexible inference with accuracy-efficiency trade-off. In this paper, by representing the topology structure of nearby users (UEs) as a graph, a multi-user channel prediction algorithm is proposed for learnable UE correlation extraction, called LUCE, to improve prediction performance. Instead of simple summation-based fusion, LUCE introduces a predefined geographic embedding and a learnable embedding to offset the position information of the cross-attention (CA) to learn and interact UE features on the graph automatically. Furthermore, a lightweight LUCE is proposed to leverage the homogeneity of CSIs by sharing subnetworks. In addition, a variable-depth learning architecture is proposed, called adaptive LUCE (AdaLUCE), and its number of blocks can be dynamically adjusted for each input by tuning a desired threshold. In particular, AdaLUCE utilizes a hierarchical triple residual architecture to promote its training efficacy and produce predictions at its internal blocks without increasing complexity, and adopts a fitting inspired criterion (FIC) to make real-time decisions for adaptive inference. In total, AdaLUCE is optimized by the weighted sum of multiple loss functions, and a multi-step training scheme is presented. We show that AdaLUCE preserves and releases computing resources for various speeds in sample-wise, improving both prediction accuracy and inference efficiency.
- 27 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 24 (8) , 6413-6428
Variable-Depth Learning Architecture to Adaptive Multi-User Channel Prediction
On the road to 6G, the growing number of antennas and mobility-related applications emphasize the urgency of addressing the channel aging issue. Traditional channel prediction methods are no longer valid due to simplified assumptions. Deep learning (DL)-based predictors, even with attractive performance improvements, generally lack the utilization of extra correlations and flexible inference with accuracy-efficiency trade-off. In this paper, by representing the topology structure of nearby users (UEs) as a graph, a multi-user channel prediction algorithm is proposed for learnable UE correlation extraction, called LUCE, to improve prediction performance. Instead of simple summation-based fusion, LUCE introduces a predefined geographic embedding and a learnable embedding to offset the position information of the cross-attention (CA) to learn and interact UE features on the graph automatically. Furthermore, a lightweight LUCE is proposed to leverage the homogeneity of CSIs by sharing subnetworks. In addition, a variable-depth learning architecture is proposed, called adaptive LUCE (AdaLUCE), and its number of blocks can be dynamically adjusted for each input by tuning a desired threshold. In particular, AdaLUCE utilizes a hierarchical triple residual architecture to promote its training efficacy and produce predictions at its internal blocks without increasing complexity, and adopts a fitting inspired criterion (FIC) to make real-time decisions for adaptive inference. In total, AdaLUCE is optimized by the weighted sum of multiple loss functions, and a multi-step training scheme is presented. We show that AdaLUCE preserves and releases computing resources for various speeds in sample-wise, improving both prediction accuracy and inference efficiency.
- 24 May 2024
- Published by Taylor & Francis in Communications in Statistics - Simulation and Computation
- Vol. 54 (9) , 3545-3572
Modified exponentially weighted moving average control chart for monitoring process dispersion
- 24 May 2024
- Published by Taylor & Francis in Communications in Statistics - Simulation and Computation
- Vol. 54 (9) , 3545-3572
Modified exponentially weighted moving average control chart for monitoring process dispersion