Personalized Federated Learning via Backbone Self-Distillation

TL;DR

This paper introduces a novel personalized federated learning method that uses backbone self-distillation, allowing clients to personalize models effectively while sharing only backbone weights for aggregation.

Contribution

It proposes a backbone self-distillation technique that enables personalized federated learning by separating shared and private model components and transferring global knowledge effectively.

Findings

Outperforms 12 state-of-the-art methods in experiments.

Effective separation of shared backbone and private head improves personalization.

Global backbone aggregation enhances local model performance.

Abstract

In practical scenarios, federated learning frequently necessitates training personalized models for each client using heterogeneous data. This paper proposes a backbone self-distillation approach to facilitate personalized federated learning. In this approach, each client trains its local model and only sends the backbone weights to the server. These weights are then aggregated to create a global backbone, which is returned to each client for updating. However, the client's local backbone lacks personalization because of the common representation. To solve this problem, each client further performs backbone self-distillation by using the global backbone as a teacher and transferring knowledge to update the local backbone. This process involves learning two components: the shared backbone for common representation and the private head for local personalization, which enables effective global knowledge transfer. Extensive experiments and comparisons with 12 state-of-the-art approaches demonstrate the effectiveness of our approach.