Federated Learning in Brain-Computer Interfaces: A Systematic Review on Aggregating Intrinsic Data for Enhanced Performance

brain-computer interfaces (BCIs)

Abstract

The development of brain-computer interfaces (BCIs) that leverage deep learning models is often hindered by the challenge of data scarcity. Despite the extensive efforts of numerous research groups and institutions in compiling EEG datasets, the variability in device usage across different sites creates significant obstacles for data sharing. In response to these challenges, this paper introduces FLEEG (Federated Learning EEG), a novel model designed to integrate data from diverse formats during the training process. FLEEG assigns each client a specific dataset and utilizes a hierarchical, personalized approach to manage these varying data formats, thereby enhancing the exchange of information. The training process is centrally coordinated by a server that aggregates knowledge from all participating datasets, ultimately improving overall performance. Experimental results demonstrate that this framework can boost classification performance by up to 84%.