Xuhua Zhao

Prof. Xuhua Zhao


Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
18 July 2024 in Electronics
Verified
Journal Article
Open access
PDF file
XZXuhua Zhao
Xuhua Zhao
CYChao Yang
Chao Yang
Donglin ZhuDonglin Zhu
Donglin Zhu
YLYujia Liu
Yujia Liu
To improve the performance of the sparrow search algorithm in solving complex optimization problems, this study proposes a novel variant called the Improved Beetle Antennae Search-Based Sparrow Search Algorithm (IBSSA). A new elite dynamic opposite learning strategy is proposed in the population initialization stage to enhance population diversity. In the update stage of the discoverer, a staged inertia weight guidance mechanism is used to improve the update formula of the discoverer, promote the information exchange between individuals, and improve the algorithm’s ability to optimize on a global level. After the follower’s position is updated, the logarithmic spiral opposition-based learning strategy is introduced to disturb the initial position of the individual in the beetle antennae search algorithm to obtain a more purposeful solution. To address the issue of decreased diversity and susceptibility to local optima in the sparrow population during later stages, the improved beetle antennae search algorithm and sparrow search algorithm are combined using a greedy strategy. This integration aims to improve convergence accuracy. On 20 benchmark test functions and the CEC2017 Test suite, IBSSA performed better than other advanced algorithms. Moreover, six engineering optimization problems were used to demonstrate the improved algorithm’s effectiveness and feasibility.

17 March 2025 in Biomimetics
Verified
Journal Article
Open access
PDF file
XZXuhua Zhao
Xuhua Zhao
YZYongming Zheng
Yongming Zheng
JWJiaxiang Wan
Jiaxiang Wan
YLYehong Li
Yehong Li
Donglin ZhuDonglin Zhu
Donglin Zhu
HLHuijuan Lu
Huijuan Lu
Federated learning (FL) is a distributed learning technique that ensures data privacy and has shown significant potential in cross-institutional image analysis. However, existing methods struggle with the inherent dynamic heterogeneity of real-world data, such as changes in cellular differentiation during disease progression or feature distribution shifts due to different imaging devices. This dynamic heterogeneity can cause catastrophic forgetting, leading to reduced performance in medical predictions across stages. Unlike previous federated learning studies that paid insufficient attention to dynamic heterogeneity, this paper proposes the FedDyH framework to address this challenge. Inspired by the adaptive regulation mechanisms of biological systems, this framework incorporates several core modules to tackle the issues arising from dynamic heterogeneity. First, the framework simulates intercellular information transfer through cross-client knowledge distillation, preserving local features while mitigating knowledge forgetting. Additionally, a dynamic regularization term is designed in which the strength can be adaptively adjusted based on real-world conditions. This mechanism resembles the role of regulatory T cells in the immune system, balancing global model convergence with local specificity adjustments to enhance the robustness of the global model while preventing interference from diverse client features. Finally, the framework introduces a genetic algorithm (GA) to simulate biological evolution, leveraging mechanisms such as gene selection, crossover, and mutation to optimize hyperparameter configurations. This enables the model to adaptively find the optimal hyperparameters in an ever-changing environment, thereby improving both adaptability and performance. Prior to this work, few studies have explored the use of optimization algorithms for hyperparameter tuning in federated learning. Experimental results demonstrate that the FedDyH framework improves accuracy compared to the SOTA baseline FedDecorr by 2.59%, 0.55%, and 5.79% on the MNIST, Fashion-MNIST, and CIFAR-10 benchmark datasets, respectively. This framework effectively addresses data heterogeneity issues in dynamic heterogeneous environments, providing an innovative solution for achieving more stable and accurate distributed federated learning.