University of Michigan Researchers Open-Source ‘FedScale’: a Federated Learning (FL) Benchmarking Suite with Realistic Datasets and a Scalable Runtime to Enable Reproducible FL Research on Privacy-Preserving Machine Learning

Federated learning (FL) is a new machine learning (ML) environment in which a logically centralized coordinator orchestrates numerous dispersed clients (e.g., cellphones or laptops) to train or assess a model collectively. It enables model training and assessment of end-user data while avoiding significant costs and privacy hazards associated with acquiring raw data from customers, with applications spanning a wide range of ML jobs. Existing work has focused on improving critical features of FL in the context of varied execution speeds …

benchmarking datasets federated learning learning machine machine learning machinelearningnews privacy research researchers scalable university university of michigan