Personalized Semantics Excitation for Federated Image Classification
Lists
Only you see this
Abstract
Federated learning casts a light on the collaboration of distributed local clients with privacy protected to attain a more generic global model. However, significant distribution shift in input/label space across different clients makes it challenging to well generalize to all clients, which motivates personalized federated learning (PFL). Existing PFL methods typically customize the local model by fine-tuning with limited local supervision and the global model regularizer, which secures local specificity but risks ruining the global discriminative knowledge. In this paper, we propose a novel Personalized Semantics Excitation ($\textbf{PSE}$) mechanism to breakthrough this limitation by exciting and fusing $\textit{personalized}$ semantics from the global model during local model customization. Specifically, PSE explores channel-wise gradient differentiation across global and local models to identify important low-level semantics mostly from convolutional layers which are embedded into the client-specific training. In addition, PSE deploys the collaboration of global and local models to enrich high-level feature representations and facilitate the robustness of client classifier through a cross-model attention module. Extensive experiments and analysis on various image classification benchmarks demonstrate the effectiveness and advantage of our method over the state-of-the-art PFL methods.
BibTeX
@inproceedings{xia2023personalized,
author = {Haifeng Xia and Kai Li and Zhengming Ding},
title = {{Personalized Semantics Excitation for Federated Image Classification}},
booktitle = {IEEE/CVF International Conference on Computer Vision},
year = {2023},
doi = {10.1109/iccv51070.2023.01768},
}Loading related content...