Learning search spaces for Bayesian optimization: Another view of hyperparameter transfer learning

TL;DR

This paper introduces a method to automatically learn and adapt the search space for Bayesian optimization using historical data, significantly improving efficiency and transfer learning capabilities in hyperparameter tuning.

Contribution

The authors propose a novel approach to automatically design the BO search space from past evaluations, enhancing transfer learning and reducing the need for manual range specification.

Findings

Boosts BO efficiency by reducing search space size

Combines with random search for parameter-free optimization

Provides a robust baseline for hyperparameter transfer learning

Abstract

Bayesian optimization (BO) is a successful methodology to optimize black-box functions that are expensive to evaluate. While traditional methods optimize each black-box function in isolation, there has been recent interest in speeding up BO by transferring knowledge across multiple related black-box functions. In this work, we introduce a method to automatically design the BO search space by relying on evaluations of previous black-box functions. We depart from the common practice of defining a set of arbitrary search ranges a priori by considering search space geometries that are learned from historical data. This simple, yet effective strategy can be used to endow many existing BO methods with transfer learning properties. Despite its simplicity, we show that our approach considerably boosts BO by reducing the size of the search space, thus accelerating the optimization of a variety of black-box optimization problems. In particular, the proposed approach combined with random search results in a parameter-free, easy-to-implement, robust hyperparameter optimization strategy. We hope it will constitute a natural baseline for further research attempting to warm-start BO.