Federated Learning with Swift: An Extension of Flower and Performance Evaluation
Maximilian Kapsecker,
Daniel N. Nugraha,
Christoph Weinhuber,
Nicholas Lane,
Stephan M. Jonas
Affiliations
Maximilian Kapsecker
TUM School of Computation, Information, and Technology, Technical University of Munich, Boltzmannstraße 3, 85748 Garching bei München, Germany; Institute for Digital Medicine, University Hospital Bonn, Venusberg-Campus 1, 53127 Bonn, Germany; Corresponding author at: TUM School of Computation, Information, and Technology, Technical University of Munich, Boltzmannstraße 3, 85748 Garching bei München, Germany.
Daniel N. Nugraha
TUM School of Computation, Information, and Technology, Technical University of Munich, Boltzmannstraße 3, 85748 Garching bei München, Germany
Christoph Weinhuber
TUM School of Computation, Information, and Technology, Technical University of Munich, Boltzmannstraße 3, 85748 Garching bei München, Germany
Nicholas Lane
Department of Computer Science and Technology, University of Cambridge, 15 JJ Thomson Ave, Cambridge CB3 0FD, United Kingdom; Flower Labs GmbH, Winterhuder Weg 29, 22085 Hamburg, Germany
Stephan M. Jonas
Institute for Digital Medicine, University Hospital Bonn, Venusberg-Campus 1, 53127 Bonn, Germany
Federated learning (FL) enables the optimization of machine learning models on distributed clients without sharing local data. The integration of FL into a mobile environment is becoming more feasible due to increasing on-device processing capabilities. However, there is limited open-source support for the iOS platform.The article introduces a Swift-based client implementation of the user-friendly FL framework Flower. The objective is facilitating FL client processes based on a modular and easy-to-integrate software development kit.A benchmark test demonstrates consistent stability and performance using the software, further motivating its use for research.
