Hongxin Zhi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Hongxin Zhi has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
Shaomei Li
3 shared publicationsHongtao Yu
2 shared publications- Hongtao Yu
Institute of Information Technology, Information Engineering University, Zhengzhou, China
2 shared publications Chao Gao
1 shared publicationFacheng Yan
1 shared publication- Hongji Liu
National Digital Switching System Engineering & Technological R&D Center, Information Engineering University, Zhengzhou 450001, China
1 shared publication
Latest Publications
- 30 January 2026
- Published by Springer Nature
- p. 339-352
CLoRA: Adversarial-Free Robustness Transfer for Self-supervised Learning
- 30 January 2026
- Published by Springer Nature
- p. 339-352
CLoRA: Adversarial-Free Robustness Transfer for Self-supervised Learning
- Open Access
- 31 July 2025
- Published by Springer Nature in Complex & Intelligent Systems
- Vol. 11 (9) , 1-18
Towards fair class-wise robustness: class optimal distribution adversarial training
- 31 July 2025
- Published by Springer Nature in Complex & Intelligent Systems
- Vol. 11 (9) , 1-18
Open AccessTowards fair class-wise robustness: class optimal distribution adversarial training
- Open Access
Towards Fair Class-wise Robustness: Class Optimal Distribution Adversarial Training
Adversarial training has proven to be a highly effective method for improving the robustness of deep neural networks against adversarial attacks. Nonetheless, it has been observed to exhibit a limitation in terms of robust fairness, characterized by a significant disparity in robustness across different classes. Recent efforts to mitigate this problem have turned to class-wise reweighted methods. However, these methods suffer from a lack of rigorous theoretical analysis and are limited in their exploration of the weight space, as they mainly rely on existing heuristic algorithms or intuition to compute weights. In addition, these methods fail to guarantee the consistency of the optimization direction due to the decoupled optimization of weights and the model parameters. They potentially lead to suboptimal weight assignments and consequently, a suboptimal model. To address these problems, this paper proposes a novel min-max training framework, Class Optimal Distribution Adversarial Training (CODAT), which employs distributionally robust optimization to fully explore the class-wise weight space, thus enabling the identification of the optimal weight with theoretical guarantees. Furthermore, we derive a closed-form optimal solution to the internal maximization and then get a deterministic equivalent objective function, which provides a theoretical basis for the joint optimization of weights and model parameters. Meanwhile, we propose a fairness elasticity coefficient for the evaluation of the algorithm with regard to both robustness and robust fairness. Experimental results on various datasets show that the proposed method can effectively improve the robust fairness of the model and outperform the state-of-the-art approaches.
Open AccessTowards Fair Class-wise Robustness: Class Optimal Distribution Adversarial Training
Adversarial training has proven to be a highly effective method for improving the robustness of deep neural networks against adversarial attacks. Nonetheless, it has been observed to exhibit a limitation in terms of robust fairness, characterized by a significant disparity in robustness across different classes. Recent efforts to mitigate this problem have turned to class-wise reweighted methods. However, these methods suffer from a lack of rigorous theoretical analysis and are limited in their exploration of the weight space, as they mainly rely on existing heuristic algorithms or intuition to compute weights. In addition, these methods fail to guarantee the consistency of the optimization direction due to the decoupled optimization of weights and the model parameters. They potentially lead to suboptimal weight assignments and consequently, a suboptimal model. To address these problems, this paper proposes a novel min-max training framework, Class Optimal Distribution Adversarial Training (CODAT), which employs distributionally robust optimization to fully explore the class-wise weight space, thus enabling the identification of the optimal weight with theoretical guarantees. Furthermore, we derive a closed-form optimal solution to the internal maximization and then get a deterministic equivalent objective function, which provides a theoretical basis for the joint optimization of weights and model parameters. Meanwhile, we propose a fairness elasticity coefficient for the evaluation of the algorithm with regard to both robustness and robust fairness. Experimental results on various datasets show that the proposed method can effectively improve the robust fairness of the model and outperform the state-of-the-art approaches.
- Open Access
- 5 February 2024
- Published by Wiley in Advanced Intelligent Systems
- Vol. 6 (5)
MA‐CAT: Misclassification‐Aware Contrastive Adversarial Training
Vulnerability to adversarial examples poses a significant challenge to the secure application of deep neural networks. Adversarial training and its variants have shown great potential in addressing this problem. However, such approaches, which directly optimize the decision boundary, often result in overly complex adversarial decision boundaries that are detrimental to generalization. To deal with this issue, a novel plug‐and‐play method known as Misclassification‐Aware Contrastive Adversarial Training (MA‐CAT) from the perspective of data distribution optimization is proposed. MA‐CAT leverages supervised decoupled contrastive learning to cluster nature examples within the same class in the logit space, indirectly increasing the margins of examples. Moreover, by taking into account the varying difficulty levels of adversarial training for different examples, MA‐CAT adaptively customizes the strength of adversarial training for each example using an instance‐wise misclassification‐aware adaptive temperature coefficient. Extensive experiments on the CIFAR‐10, CIFAR‐100, and SVHN datasets demonstrate that MA‐CAT can be easily integrated into existing models and significantly improves robustness with minimal computational cost.
- 5 February 2024
- Published by Wiley in Advanced Intelligent Systems
- Vol. 6 (5)
Open AccessMA‐CAT: Misclassification‐Aware Contrastive Adversarial Training
Vulnerability to adversarial examples poses a significant challenge to the secure application of deep neural networks. Adversarial training and its variants have shown great potential in addressing this problem. However, such approaches, which directly optimize the decision boundary, often result in overly complex adversarial decision boundaries that are detrimental to generalization. To deal with this issue, a novel plug‐and‐play method known as Misclassification‐Aware Contrastive Adversarial Training (MA‐CAT) from the perspective of data distribution optimization is proposed. MA‐CAT leverages supervised decoupled contrastive learning to cluster nature examples within the same class in the logit space, indirectly increasing the margins of examples. Moreover, by taking into account the varying difficulty levels of adversarial training for different examples, MA‐CAT adaptively customizes the strength of adversarial training for each example using an instance‐wise misclassification‐aware adaptive temperature coefficient. Extensive experiments on the CIFAR‐10, CIFAR‐100, and SVHN datasets demonstrate that MA‐CAT can be easily integrated into existing models and significantly improves robustness with minimal computational cost.
- Open Access
- 18 April 2023
- Published by MDPI AG in Electronics
- Vol. 12 (8) , 1905
Multi-Hop Knowledge Graph Question Answer Method Based on Relation Knowledge Enhancement
Multi-hop knowledge graph question answer (KGQA) is a challenging task because it requires reasoning over multiple edges of the knowledge graph (KG) to arrive at the right answer. However, KGs are often incomplete with many missing links, posing additional challenges for multi-hop KGQA. Recent research on multi-hop KGQA attempted to deal with KG sparsity with relevant external texts. In our work, we propose a multi-hop KGQA model based on relation knowledge enhancement (RKE-KGQA), which fuses both label and text relations through global attention for relation knowledge augmentation. It is well known that the relation between entities can be represented by labels in the knowledge graph or texts in the text corpus, and multi-hop KGQA needs to jump across different entities through relations. First, we assign an activation probability to each entity, then calculate a score for the enhancement relation, and then transfer the score through the activated relations and, finally, obtain the answer. We carry out extensive experiments on three datasets and demonstrate that RKE-KGQA achieves the outperformance result.
- 18 April 2023
- Published by MDPI AG in Electronics
- Vol. 12 (8) , 1905
Open AccessMulti-Hop Knowledge Graph Question Answer Method Based on Relation Knowledge Enhancement
Multi-hop knowledge graph question answer (KGQA) is a challenging task because it requires reasoning over multiple edges of the knowledge graph (KG) to arrive at the right answer. However, KGs are often incomplete with many missing links, posing additional challenges for multi-hop KGQA. Recent research on multi-hop KGQA attempted to deal with KG sparsity with relevant external texts. In our work, we propose a multi-hop KGQA model based on relation knowledge enhancement (RKE-KGQA), which fuses both label and text relations through global attention for relation knowledge augmentation. It is well known that the relation between entities can be represented by labels in the knowledge graph or texts in the text corpus, and multi-hop KGQA needs to jump across different entities through relations. First, we assign an activation probability to each entity, then calculate a score for the enhancement relation, and then transfer the score through the activated relations and, finally, obtain the answer. We carry out extensive experiments on three datasets and demonstrate that RKE-KGQA achieves the outperformance result.