Abstract

Very few methods for hybrid federated learning, where clients only holdsubsets of both features and samples, exist. Yet, this scenario is extremelyimportant in practical settings. We provide a fast, robust algorithm for hybridfederated learning that hinges on Fenchel Duality. We prove the convergence ofthe algorithm to the same solution as if the model is trained centrally in avariety of practical regimes. Furthermore, we provide experimental results thatdemonstrate the performance improvements of the algorithm over a commonly usedmethod in federated learning, FedAvg, and an existing hybrid FL algorithm,HyFEM. We also provide privacy considerations and necessary steps to protectclient data.

 

Quick Read (beta)