FedSPC: Shared Parameter Correction for Personalized Federated Learning
FedSPC: Shared Parameter Correction for Personalized Federated Learning
FedSPC：用于个性化联邦学习的共享参数校正
Abstract: Personalized federated learning (PFL) is one of the important approaches in federated learning for addressing statistical heterogeneity while enabling client-specific adaptation. Many PFL methods split the model into shared and personalized parameters, which are jointly trained on each client. However, this creates an optimization issue: shared parameters are updated by clients optimizing different local objectives, which can lead to inconsistent shared updates and weaken the shared representation.
摘要： 个性化联邦学习（PFL）是联邦学习中解决统计异构性并实现客户端特定适应的重要方法之一。许多 PFL 方法将模型拆分为共享参数和个性化参数，并在每个客户端上进行联合训练。然而，这带来了一个优化问题：共享参数由优化不同局部目标的客户端进行更新，这可能导致共享更新的不一致，并削弱共享表示的效果。
To address this problem, we propose Federated Shared Parameter Correction (FedSPC), a modular correction method for PFL. FedSPC applies control-variate correction only to the shared parameters of a given PFL method, while leaving personalized parameters unchanged. It can be integrated into three common PFL settings: shared feature extractors, shared classifiers, and fully shared models with local regularization.
为了解决这一问题，我们提出了联邦共享参数校正（FedSPC），这是一种用于 PFL 的模块化校正方法。FedSPC 仅对给定 PFL 方法的共享参数应用控制变量校正，同时保持个性化参数不变。它可以集成到三种常见的 PFL 设置中：共享特征提取器、共享分类器以及带有局部正则化的全共享模型。
Experiments on CIFAR-100 and Tiny-ImageNet with ViT, ResNet-34, and VGG-11 show that FedSPC improves performance across representative PFL methods, including FedPer, FedRep, FedBABU, LG-FedAvg, and Ditto.
在 CIFAR-100 和 Tiny-ImageNet 数据集上，使用 ViT、ResNet-34 和 VGG-11 模型进行的实验表明，FedSPC 提升了多种代表性 PFL 方法的性能，包括 FedPer、FedRep、FedBABU、LG-FedAvg 和 Ditto。
Paper Details:
- Authors: Kannanthodath Induchoodan Ajay Menon, Christian Prehofer, Yunfei Xu, Toru Hirano
- arXiv ID: 2606.13748
- Subject: Machine Learning (cs.LG)
论文详情：
- 作者： Kannanthodath Induchoodan Ajay Menon, Christian Prehofer, Yunfei Xu, Toru Hirano
- arXiv ID: 2606.13748
- 学科： 机器学习 (cs.LG)