Exploiting Label Skews in Federated Learning with Model Concatenation

TL;DR

This paper introduces FedConcat, a novel federated learning method that concatenates local models instead of averaging, effectively handling label skew non-IID data, improving accuracy, and reducing communication costs.

Contribution

FedConcat is a new approach that concatenates local models and uses clustering to improve federated learning under label skew non-IID data.

Findings

FedConcat outperforms state-of-the-art methods in accuracy.

It reduces communication costs compared to existing approaches.

Clustering clients by label distribution enhances model performance.

Abstract

Federated Learning (FL) has emerged as a promising solution to perform deep learning on different data owners without exchanging raw data. However, non-IID data has been a key challenge in FL, which could significantly degrade the accuracy of the final model. Among different non-IID types, label skews have been challenging and common in image classification and other tasks. Instead of averaging the local models in most previous studies, we propose FedConcat, a simple and effective approach that concatenates these local models as the base of the global model to effectively aggregate the local knowledge. To reduce the size of the global model, we adopt the clustering technique to group the clients by their label distributions and collaboratively train a model inside each cluster. We theoretically analyze the advantage of concatenation over averaging by analyzing the information bottleneck of deep neural networks. Experimental results demonstrate that FedConcat achieves significantly higher accuracy than previous state-of-the-art FL methods in various heterogeneous label skew distribution settings and meanwhile has lower communication costs. Our code is publicly available at https://github.com/sjtudyq/FedConcat.