Privately Customizing Prefinetuning to Better Match User Data in Federated Learning

TL;DR

This paper introduces FreD, a privacy-preserving method to evaluate and select prefinetuning datasets in federated learning, enhancing model customization to user data while maintaining privacy.

Contribution

The paper proposes FreD, a novel differentially-private Fréchet distance measure for evaluating prefinetuning datasets in federated learning, improving dataset selection accuracy.

Findings

FreD accurately predicts the best prefinetuning dataset.

FreD operates with minimal privacy cost.

The approach enables better dataset customization in federated learning.

Abstract

In Federated Learning (FL), accessing private client data incurs communication and privacy costs. As a result, FL deployments commonly prefinetune pretrained foundation models on a (large, possibly public) dataset that is held by the central server; they then FL-finetune the model on a private, federated dataset held by clients. Evaluating prefinetuning dataset quality reliably and privately is therefore of high importance. To this end, we propose FreD (Federated Private Fr\'echet Distance) -- a privately computed distance between a prefinetuning dataset and federated datasets. Intuitively, it privately computes and compares a Fr\'echet distance between embeddings generated by a large language model on both the central (public) dataset and the federated private client data. To make this computation privacy-preserving, we use distributed, differentially-private mean and covariance estimators. We show empirically that FreD accurately predicts the best prefinetuning dataset at minimal privacy cost. Altogether, using FreD we demonstrate a proof-of-concept for a new approach in private FL training: (1) customize a prefinetuning dataset to better match user data (2) prefinetune (3) perform FL-finetuning.