ZooPFL: Exploring Black-box Foundation Models for Personalized Federated Learning

TL;DR

This paper introduces ZOOPFL, a method that enables personalized federated learning with black-box foundation models by using zeroth-order optimization and input adaptation to address resource limitations and distribution shifts.

Contribution

It proposes a novel zeroth-order optimization approach for FL that avoids direct model access and incorporates input surgery and low-dimensional embeddings for personalization and efficiency.

Findings

Effective personalization on vision and NLP tasks

Reduces computation costs with input surgery

Achieves convergence with theoretical guarantees

Abstract

When personalized federated learning (FL) meets large foundation models, new challenges arise from various limitations in resources. In addition to typical limitations such as data, computation, and communication costs, access to the models is also often limited. This paper endeavors to solve both the challenges of limited resources and personalization. i.e., distribution shifts between clients. To do so, we propose a method named ZOOPFL that uses Zeroth-Order Optimization for Personalized Federated Learning. ZOOPFL avoids direct interference with the foundation models and instead learns to adapt its inputs through zeroth-order optimization. In addition, we employ simple yet effective linear projections to remap its predictions for personalization. To reduce the computation costs and enhance personalization, we propose input surgery to incorporate an auto-encoder with low-dimensional and client-specific embeddings. We provide theoretical support for ZOOPFL to analyze its convergence. Extensive empirical experiments on computer vision and natural language processing tasks using popular foundation models demonstrate its effectiveness for FL on black-box foundation models.