FACT: Federated Adversarial Cross Training

TL;DR

FACT introduces a federated learning method that leverages domain differences among clients to improve model adaptation to new, unlabeled target domains, outperforming existing models across multiple benchmarks.

Contribution

The paper presents a novel federated adversarial training approach that exploits source client domain differences for better target domain adaptation without labeled data.

Findings

Outperforms state-of-the-art federated and domain adaptation models

Effective under communication restrictions and varying client numbers

Achieves superior results on multiple benchmarks

Abstract

Federated Learning (FL) facilitates distributed model development to aggregate multiple confidential data sources. The information transfer among clients can be compromised by distributional differences, i.e., by non-i.i.d. data. A particularly challenging scenario is the federated model adaptation to a target client without access to annotated data. We propose Federated Adversarial Cross Training (FACT), which uses the implicit domain differences between source clients to identify domain shifts in the target domain. In each round of FL, FACT cross initializes a pair of source clients to generate domain specialized representations which are then used as a direct adversary to learn a domain invariant data representation. We empirically show that FACT outperforms state-of-the-art federated, non-federated and source-free domain adaptation models on three popular multi-source-single-target benchmarks, and state-of-the-art Unsupervised Domain Adaptation (UDA) models on single-source-single-target experiments. We further study FACT's behavior with respect to communication restrictions and the number of participating clients.