On Noisy Evaluation in Federated Hyperparameter Tuning

TL;DR

This paper systematically studies the impact of noise on federated hyperparameter tuning, revealing significant effects even from small noise levels, and proposes using public proxy data to improve evaluation accuracy.

Contribution

It is the first comprehensive analysis of noisy evaluation in federated hyperparameter tuning and introduces a proxy data method to mitigate noise effects.

Findings

Small noise significantly degrades tuning performance

State-of-the-art methods are as ineffective as naive baselines under noise

Using public proxy data improves evaluation reliability

Abstract

Hyperparameter tuning is critical to the success of federated learning applications. Unfortunately, appropriately selecting hyperparameters is challenging in federated networks. Issues of scale, privacy, and heterogeneity introduce noise in the tuning process and make it difficult to evaluate the performance of various hyperparameters. In this work, we perform the first systematic study on the effect of noisy evaluation in federated hyperparameter tuning. We first identify and rigorously explore key sources of noise, including client subsampling, data and systems heterogeneity, and data privacy. Surprisingly, our results indicate that even small amounts of noise can significantly impact tuning methods-reducing the performance of state-of-the-art approaches to that of naive baselines. To address noisy evaluation in such scenarios, we propose a simple and effective approach that leverages public proxy data to boost the evaluation signal. Our work establishes general challenges, baselines, and best practices for future work in federated hyperparameter tuning.