Code supporting the publication: Nomad: Accelerating Geo-distributed Learning with Client Transfers
What's in it
This repository contains the source code and simulation framework used in the publication "Nomad: Accelerating Geo-distributed Learning with Client Transfers" . The framework enables experimentation with asynchronous multi-server Federated Learning (FL) in geo-distributed environments, supporting dynamic client-to-server reassignment, multiple aggregation strategies, heterogeneous client data distributions, varying network latencies, and client churn scenarios. Nomad implements adaptive client transfer mechanisms that optimize server assignments based on both communication latency and data distribution characteristics, facilitating the evaluation of training efficiency, convergence behavior, and model accuracy under realistic geo-distributed FL settings.
Files and services
- DataCite REST API (JSON)api.datacite.orgOpen