Improving Federated Learning Personalization via Model Agnostic Meta Learning

TL;DR

This paper explores how Model Agnostic Meta Learning (MAML) can enhance personalization in Federated Learning by interpreting Federated Averaging as a meta learning algorithm and analyzing its implications for model fine-tuning and personalization.

Contribution

It demonstrates the connection between Federated Averaging and meta learning, and shows how fine-tuning can improve personalization while highlighting the limitations of global accuracy optimization.

Findings

Federated Averaging can be viewed as a meta learning algorithm.

Careful fine-tuning improves personalization and model accuracy.

Models trained with federated methods are easier to personalize than those trained centrally.

Abstract

Federated Learning (FL) refers to learning a high quality global model based on decentralized data storage, without ever copying the raw data. A natural scenario arises with data created on mobile phones by the activity of their users. Given the typical data heterogeneity in such situations, it is natural to ask how can the global model be personalized for every such device, individually. In this work, we point out that the setting of Model Agnostic Meta Learning (MAML), where one optimizes for a fast, gradient-based, few-shot adaptation to a heterogeneous distribution of tasks, has a number of similarities with the objective of personalization for FL. We present FL as a natural source of practical applications for MAML algorithms, and make the following observations. 1) The popular FL algorithm, Federated Averaging, can be interpreted as a meta learning algorithm. 2) Careful fine-tuning can yield a global model with higher accuracy, which is at the same time easier to personalize. However, solely optimizing for the global model accuracy yields a weaker personalization result. 3) A model trained using a standard datacenter optimization method is much harder to personalize, compared to one trained using Federated Averaging, supporting the first claim. These results raise new questions for FL, MAML, and broader ML research.