EFFGAN: Ensembles of fine-tuned federated GANs

Decentralized machine learning tackles the problem of learning useful models when data is distributed among several clients. The most prevalent decentralized setting today is federated learning (FL), where a central server orchestrates the learning among clients. In this work, we contribute to the relatively understudied sub-field of generative modelling in the FL framework.We study the task of how to train generative adversarial networks (GANs) when training data is heterogeneously distributed (non-iid) over clients and cannot be shared. Our objective is to train a generator that is able to sample from the collective data distribution centrally, while the client data never leaves the clients and user privacy is respected. We show using standard benchmark image datasets that existing approaches fail in this setting, experiencing so-called client drift when the local number of epochs becomes to large and local parameters drift too far away in parameter space. To tackle this challenge, we propose a novel approach named EFFGAN: Ensembles of fine-tuned federated GANs. Being an ensemble of local expert generators, EFFGAN is able to learn the data distribution over all clients and mitigate client drift. It is able to train with a large number of local epochs, making it more communication efficient than previous works.