论文概要
研究领域: CV 作者: Sina Gholami, Abdulmoneam Ali, Tania Haghighi 发布时间: 2026-04-22 arXiv: 2604.20825
中文摘要
联邦学习（FL）支持在不共享原始数据的情况下协作训练模型；然而，分布式客户端中存在的噪声标签可能严重降低学习性能。本文提出FedSIR，一种面向噪声标签的鲁棒联邦学习多阶段框架。与现有主要依赖设计噪声容忍损失函数或利用训练过程中损失动态的方法不同，我们的方法利用客户端特征表示的谱结构来识别和缓解标签噪声。框架包含三个关键组件。首先，我们通过分析类别特征子空间的谱一致性，以最小通信开销识别干净和噪声客户端。其次，干净客户端提供谱参考，使噪声客户端能够利用主导类别方向和残差子空间对潜在损坏样本进行重标签。第三，我们采用噪声感知训练策略，整合logit调整损失、知识蒸馏和距离感知聚合，以进一步稳定联邦优化。在标准FL基准上的大量实验表明，FedSIR在面向噪声标签的联邦学习中持续优于最先进的方法。
原文摘要
Federated learning (FL) enables collaborative model training without sharing raw data; however, the presence of noisy labels across distributed clients can severely degrade the learning performance. In this paper, we propose FedSIR, a multi-stage framework for robust FL under noisy labels. Different from existing approaches that mainly rely on designing noise-tolerant loss functions or exploiting loss dynamics during training, our method leverages the spectral structure of client feature representations to identify and mitigate label noise. Our framework consists of three key components. First, we identify clean and noisy clients by analyzing the spectral consistency of class-wise feature subspaces with minimal communication overhead. Second, clean clients provide spectral references that ...
--- *自动采集于 2026-04-24*
#论文 #arXiv #CV #小凯