Jointly Learning from Decentralized (Federated) and Centralized Data to Mitigate Distribution Shift

TL;DR

This paper proposes a method to combine federated learning with centralized datacenter data to better address distribution shifts, improving model performance while respecting privacy constraints.

Contribution

It introduces a novel approach to mix decentralized and centralized data during training to mitigate distribution shift in federated learning settings.

Findings

Improved model accuracy with mixed data training.

Effective mitigation of distribution shift.

Maintained privacy constraints during training.

Abstract

With privacy as a motivation, Federated Learning (FL) is an increasingly used paradigm where learning takes place collectively on edge devices, each with a cache of user-generated training examples that remain resident on the local device. These on-device training examples are gathered in situ during the course of users' interactions with their devices, and thus are highly reflective of at least part of the inference data distribution. Yet a distribution shift may still exist; the on-device training examples may lack for some data inputs expected to be encountered at inference time. This paper proposes a way to mitigate this shift: selective usage of datacenter data, mixed in with FL. By mixing decentralized (federated) and centralized (datacenter) data, we can form an effective training data distribution that better matches the inference data distribution, resulting in more useful models while still meeting the private training data access constraints imposed by FL.