FedGreen: Carbon-aware Federated Learning with Model Size Adaptation

arXiv:2404.15503v1 Announce Type: new
Abstract: Federated learning (FL) provides a promising collaborative framework to build a model from distributed clients, and this work investigates the carbon emission of the FL process. Cloud and edge servers hosting FL clients may exhibit diverse carbon footprints influenced by their geographical locations with varying power sources, offering opportunities to reduce carbon emissions by training local models with adaptive computations and communications. In this paper, we propose FedGreen, a carbon-aware FL approach to efficiently train …

abstract arxiv build carbon cloud collaborative cs.ai cs.dc cs.lg distributed diverse edge edge servers federated learning framework hosting locations power process servers type work