A Personalized Federated Learning Algorithm: an Application in Anomaly Detection

TL;DR

This paper introduces a personalized federated learning algorithm, PC-FedAvg, designed to improve model accuracy in non-IID data scenarios, demonstrated through anomaly detection tasks.

Contribution

The paper proposes a novel personalized federated learning algorithm, PC-FedAvg, that enhances model accuracy by addressing data heterogeneity in federated settings.

Findings

PC-FedAvg outperforms existing methods in accuracy.

The algorithm effectively personalizes models for diverse clients.

Experimental results validate improved generalization.

Abstract

Federated Learning (FL) has recently emerged as a promising method that employs a distributed learning model structure to overcome data privacy and transmission issues paused by central machine learning models. In FL, datasets collected from different devices or sensors are used to train local models (clients) each of which shares its learning with a centralized model (server). However, this distributed learning approach presents unique learning challenges as the data used at local clients can be non-IID (Independent and Identically Distributed) and statistically diverse which decrease learning accuracy in the central model. In this paper, we overcome this problem by proposing a novel Personalized Conditional FedAvg (PC-FedAvg) which aims to control weights communication and aggregation augmented with a tailored learning algorithm to personalize the resulting models at each client. Our experimental validation on two datasets showed that our PC-FedAvg precisely constructed generalized clients' models and thus achieved higher accuracy compared to other state-of-the-art methods.