Put CASH on Bandits: A Max K-Armed Problem for Automated Machine Learning

TL;DR

This paper introduces MaxUCB, a bandit algorithm tailored for AutoML's CASH problem, effectively balancing model selection and hyperparameter tuning with improved efficiency and performance.

Contribution

It presents a novel max k-armed bandit method designed for AutoML, specifically addressing the reward distribution characteristics in this domain.

Findings

MaxUCB outperforms prior methods on AutoML benchmarks.

The method is theoretically sound for light-tailed reward distributions.

Empirical results show improved resource allocation efficiency.

Abstract

The Combined Algorithm Selection and Hyperparameter optimization (CASH) is a challenging resource allocation problem in the field of AutoML. We propose MaxUCB, a max k-armed bandit method to trade off exploring different model classes and conducting hyperparameter optimization. MaxUCB is specifically designed for the light-tailed and bounded reward distributions arising in this setting and, thus, provides an efficient alternative compared to classic max k-armed bandit methods assuming heavy-tailed reward distributions. We theoretically and empirically evaluate our method on four standard AutoML benchmarks, demonstrating superior performance over prior approaches. We make our code and data available at https://github.com/amirbalef/CASH_with_Bandits