Your Data, My Model: Learning Who Really Helps in Federated Learning

TL;DR

This paper introduces a privacy-preserving, model-agnostic peer selection method for federated learning that identifies beneficial collaborators based on model improvement after a single gradient step, enhancing personalized model training.

Contribution

It presents a novel, simple, and privacy-preserving peer selection technique that extends to non-parametric models, improving collaboration in personalized federated learning.

Findings

Effective peer selection without sharing raw data

Applicable to both parametric and non-parametric models

Enhances personalized federated learning performance

Abstract

Many important machine learning applications involve networks of devices-such as wearables or smartphones-that generate local data and train personalized models. A key challenge is determining which peers are most beneficial for collaboration. We propose a simple and privacy-preserving method to select relevant collaborators by evaluating how much a model improves after a single gradient step using another devices data-without sharing raw data. This method naturally extends to non-parametric models by replacing the gradient step with a non-parametric generalization. Our approach enables model-agnostic, data-driven peer selection for personalized federated learning (PersFL).