Federated User Representation Learning

TL;DR

FURL introduces a federated learning approach for user personalization that preserves privacy, improves memory efficiency, and achieves comparable performance to centralized models by splitting model parameters into federated and private sets.

Contribution

The paper proposes a novel federated learning method that separates private user embeddings from shared parameters, enhancing privacy and efficiency while maintaining model quality.

Findings

FURL achieves 8% and 51% performance improvements on two datasets.

FURL maintains similar performance to centralized training with minimal accuracy loss.

User embeddings learned via FURL are structurally similar to those from centralized training.

Abstract

Collaborative personalization, such as through learned user representations (embeddings), can improve the prediction accuracy of neural-network-based models significantly. We propose Federated User Representation Learning (FURL), a simple, scalable, privacy-preserving and resource-efficient way to utilize existing neural personalization techniques in the Federated Learning (FL) setting. FURL divides model parameters into federated and private parameters. Private parameters, such as private user embeddings, are trained locally, but unlike federated parameters, they are not transferred to or averaged on the server. We show theoretically that this parameter split does not affect training for most model personalization approaches. Storing user embeddings locally not only preserves user privacy, but also improves memory locality of personalization compared to on-server training. We evaluate FURL on two datasets, demonstrating a significant improvement in model quality with 8% and 51% performance increases, and approximately the same level of performance as centralized training with only 0% and 4% reductions. Furthermore, we show that user embeddings learned in FL and the centralized setting have a very similar structure, indicating that FURL can learn collaboratively through the shared parameters while preserving user privacy.