DOI number:10.1109/OJCOMS.2026.3703734
Affiliation of Author(s):杭州国际创新研究院
Journal:IEEE Open Journal of the Communications Society
Place of Publication:美国
Key Words:federated learning; explainable artificial intelligence; data heterogeneity; robust aggregation; poisoning attacks
Abstract:提出基于可解释人工智能的鲁棒联邦学习框架XAI-FL。服务器利用少量验证数据和积分梯度评估各客户端局部模型的贡献，据此动态调整聚合权重，以缓解非独立同分布数据、恶意客户端和模型投毒对全局模型精度与稳定性的影响。
Note:2026年Early Access论文；卷7，页7103-7117。
Co-author:Chuan Ma, Youjia Chen, Hao Yu, Athanasios V. Vasilakos, Youlong Wu
First Author:Junnan Yang
Indexed by:Journal paper
Document Code:11563834
Discipline:工学
First-Level Discipline:Information and Communication Engineering
Document Type:SCI
Volume:7
Page Number:7103-7117
ISSN No.:2644-125X
Translation or Not:no
Date of Publication:2026-06-15