Enabling efficient and low-effort decentralized federated learning with the EdgeFL framework
Year: 2024
Type: article
Abstract: Federated Learning (FL) has emerged as a pivotal solution in the field of machine learning, providing a solid way to address the growing concerns about data privacy. As data collecting and analysis become more advanced across industries, the necessity to protect sensitive information and comply with data protection standards becomes increasingly important. However, despite the promise of FL, existing FL frameworks have presented substantial challenges for software engineers. These challenges include complexity in implementation, limited customization options, and scalability constraints. This has prevented the widespread adoption of FL as a data privacy-conscious machine learning solution.In response to these limitations, we introduce EdgeFL, an efficient and low-effort decentralized FL framework. EdgeFL aims to resolve the inherent constraints of centralized aggregation and scalability encountered in FL deployments. Our contribution, as demonstrated through experiments and a case study, showcases the improved learning efficiency and performance of EdgeFL in comparison to existing FL frameworks. EdgeFL achieves this by reducing weight update latency and enabling faster model evolution on edge devices.In essence, EdgeFL offers a transformative solution for software engineers and companies seeking the benefits of Federated Learning, while effectively overcoming the challenges and privacy concerns associated with traditional FL frameworks.
Institutions: Chalmers University of Technology, Malmö University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $3,480
Funder: Vetenskapsrådet
Award: NAISS
OpenAlex ID: W4403499895