Yixuan Cai
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yixuan Cai has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Jianbo Xu
Hainan University, Renmin Road, Haikou, 570-228, Hainan, China
1 shared publication - Jiantao Xu
Graduate School of Computer Science and Engineering, The University of Aizu, Aizuwakamatsu 965-8580, Fukushima Prefecture, Japan
1 shared publication - Kei Chi Wing Brian
The Hong Kong Polytechnic University
1 shared publication - Menglong Feng
College of Mechanical and Electronic Engineering, Nanjing Forestry University, Nanjing 210037, China
1 shared publication - Shen Yan
Nanjing Forestry University
1 shared publication - Yuxing Li
School of Computer Science and Technology, The University of Hainan, Haikou 570228, China
1 shared publication
Latest Publications
- Open Access
A Reputation-Aware Defense Framework for Strategic Behaviors in Federated Learning
Federated Learning (FL) enables privacy-preserving model training across distributed clients. However, its reliance on voluntary client participation makes it vulnerable to strategic behaviors—actions that are not overtly malicious but significantly impair model convergence and fairness. Existing defense methods primarily focus on explicit attacks, overlooking the challenges posed by economically motivated “pseudo-honest” clients. To address this gap, we propose a Reputation-Aware Defense Framework to mitigate strategic behaviors in FL. This framework introduces a multi-dimensional dynamic reputation model that evaluates client behaviors based on gradient alignment, participation consistency, and update stability. The resulting reputation scores are incorporated into both aggregation and incentive mechanisms, forming a behavior-feedback loop that rewards honest participation and penalizes opportunistic strategies. We theoretically prove the convergence of reputation scores, the suppression of low-quality updates in aggregation, and the emergence of honest participation as a Nash equilibrium under the incentive mechanism. Experiments on datasets such as CIFAR-10, FEMNIST, MIMIC-III demonstrate that our approach significantly outperforms baseline methods in accuracy, fairness, and robustness, even when up to 60% of clients act strategically. This study bridges trust modeling and robust optimization in FL, offering a secure foundation for federated systems operating in open and incentive-driven environments.
Open AccessA Reputation-Aware Defense Framework for Strategic Behaviors in Federated Learning
Federated Learning (FL) enables privacy-preserving model training across distributed clients. However, its reliance on voluntary client participation makes it vulnerable to strategic behaviors—actions that are not overtly malicious but significantly impair model convergence and fairness. Existing defense methods primarily focus on explicit attacks, overlooking the challenges posed by economically motivated “pseudo-honest” clients. To address this gap, we propose a Reputation-Aware Defense Framework to mitigate strategic behaviors in FL. This framework introduces a multi-dimensional dynamic reputation model that evaluates client behaviors based on gradient alignment, participation consistency, and update stability. The resulting reputation scores are incorporated into both aggregation and incentive mechanisms, forming a behavior-feedback loop that rewards honest participation and penalizes opportunistic strategies. We theoretically prove the convergence of reputation scores, the suppression of low-quality updates in aggregation, and the emergence of honest participation as a Nash equilibrium under the incentive mechanism. Experiments on datasets such as CIFAR-10, FEMNIST, MIMIC-III demonstrate that our approach significantly outperforms baseline methods in accuracy, fairness, and robustness, even when up to 60% of clients act strategically. This study bridges trust modeling and robust optimization in FL, offering a secure foundation for federated systems operating in open and incentive-driven environments.
- Open Access
- 9 May 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (10) , 1557
Enhanced ResNet50 for Diabetic Retinopathy Classification: External Attention and Modified Residual Branch
One of the common microvascular complications in diabetic patients is diabetic retinopathy (DR), which primarily impacts the retinal blood vessels. As the course of diabetes progresses, the incidence of DR gradually increases, and, in serious situations, it can cause vision loss and even blindness. Diagnosing DR early is essential to mitigate its consequences, and deep learning models provide an effective approach. In this study, we propose an improved ResNet50 model, which replaces the 3 × 3 convolution in the residual structure by introducing an external attention mechanism, which improves the model’s awareness of global information and allows the model to grasp the characteristics of the input data more thoroughly. In addition, multiscale convolution is added to the residual branch, which further improves the ability of the model to extract local features and global features, and improves the processing accuracy of image details. In addition, the Sophia optimizer is introduced to replace the traditional Adam optimizer, which further optimizes the classification performance of the model. In this study, 3662 images from the Kaggle open dataset were used to generate 20,184 images for model training after image preprocessing and data augmentation. Experimental results show that the improved ResNet50 model achieves a classification accuracy of 96.68% on the validation set, which is 4.36% higher than the original architecture, and the Kappa value is increased by 5.45%. These improvements contribute to the early diagnosis of DR and decrease the likelihood of blindness among patients.
- 9 May 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (10) , 1557
Open AccessEnhanced ResNet50 for Diabetic Retinopathy Classification: External Attention and Modified Residual Branch
One of the common microvascular complications in diabetic patients is diabetic retinopathy (DR), which primarily impacts the retinal blood vessels. As the course of diabetes progresses, the incidence of DR gradually increases, and, in serious situations, it can cause vision loss and even blindness. Diagnosing DR early is essential to mitigate its consequences, and deep learning models provide an effective approach. In this study, we propose an improved ResNet50 model, which replaces the 3 × 3 convolution in the residual structure by introducing an external attention mechanism, which improves the model’s awareness of global information and allows the model to grasp the characteristics of the input data more thoroughly. In addition, multiscale convolution is added to the residual branch, which further improves the ability of the model to extract local features and global features, and improves the processing accuracy of image details. In addition, the Sophia optimizer is introduced to replace the traditional Adam optimizer, which further optimizes the classification performance of the model. In this study, 3662 images from the Kaggle open dataset were used to generate 20,184 images for model training after image preprocessing and data augmentation. Experimental results show that the improved ResNet50 model achieves a classification accuracy of 96.68% on the validation set, which is 4.36% higher than the original architecture, and the Kappa value is increased by 5.45%. These improvements contribute to the early diagnosis of DR and decrease the likelihood of blindness among patients.