Non-stochastic Best Arm Identification and Hyperparameter Optimization

TL;DR

This paper introduces a non-stochastic best-arm identification framework tailored for hyperparameter optimization, analyzing an existing algorithm and demonstrating its efficiency in faster convergence to optimal hyperparameters.

Contribution

It extends the best-arm identification framework to the non-stochastic setting and applies it to hyperparameter optimization, providing theoretical analysis and empirical validation.

Findings

Achieves comparable test accuracies faster than baseline methods

Allocates resources more efficiently to promising hyperparameters

Demonstrates effectiveness of non-stochastic approach in hyperparameter tuning

Abstract

Motivated by the task of hyperparameter optimization, we introduce the non-stochastic best-arm identification problem. Within the multi-armed bandit literature, the cumulative regret objective enjoys algorithms and analyses for both the non-stochastic and stochastic settings while to the best of our knowledge, the best-arm identification framework has only been considered in the stochastic setting. We introduce the non-stochastic setting under this framework, identify a known algorithm that is well-suited for this setting, and analyze its behavior. Next, by leveraging the iterative nature of standard machine learning algorithms, we cast hyperparameter optimization as an instance of non-stochastic best-arm identification, and empirically evaluate our proposed algorithm on this task. Our empirical results show that, by allocating more resources to promising hyperparameter settings, we typically achieve comparable test accuracies an order of magnitude faster than baseline methods.