Federated learning with Tsetlin machine

Abstract

Federated Learning (FL) has gained significant traction in recent years across diverse domains like Automatic Speech Recognition (ASR) and Image Recognition [1]. FL addresses the critical issue of privacy, particularly concerning user data like images and personal information. FL tackles this challenge by allowing data to remain decentralized on client devices rather than being centralized on a server for training. With the increasing number of network-connected devices, including smartphones and Internet-of-Things (IoT) devices, there is a growing need for FL models that are not only communication-efficient but also capable of accommodating edge devices’ constraints. Additionally, such models must prioritize data privacy while ensuring robust learning outcomes. In FL, Deep Neural Networks (DNN) are typically employed, requiring intensive arithmetic computations for gradient descent. This involves multiple iterations of learning on individual client devices, followed by the transmission of DNN parameters across the network for aggregation on a central server [2]. Given the resource-intensive nature of DNNs, existing FL methodologies may struggle to align with the vision of deploying FL on edge devices. Compounded by challenges such as high communication overheads, data heterogeneity, and scalability, FL at the edge remains challenging [1].

More information

Identifiers

ORCID for Geoff Merrett: orcid.org/0000-0003-4980-3894

ORCID for Jonathon Hare: orcid.org/0000-0003-2921-4283

Catalogue record

Export record