Use of static surrogates in hyperparameter optimization

TL;DR

This paper enhances the HyperNOMAD hyperparameter optimization library by integrating static surrogates to efficiently discard poor configurations early, reducing resource use while maintaining solution quality.

Contribution

It introduces static surrogate-based strategies into HyperNOMAD for early stopping and ranking, improving efficiency in neural network hyperparameter tuning.

Findings

Resource consumption is reduced without quality loss.

Static surrogates effectively identify low-quality configurations.

Enhancements improve HyperNOMAD's efficiency in hyperparameter optimization.

Abstract

Optimizing the hyperparameters and architecture of a neural network is a long yet necessary phase in the development of any new application. This consuming process can benefit from the elaboration of strategies designed to quickly discard low quality configurations and focus on more promising candidates. This work aims at enhancing HyperNOMAD, a library that adapts a direct search derivative-free optimization algorithm to tune both the architecture and the training of a neural network simultaneously, by targeting two keys steps of its execution and exploiting cheap approximations in the form of static surrogates to trigger the early stopping of the evaluation of a configuration and the ranking of pools of candidates. These additions to HyperNOMAD are shown to improve on its resources consumption without harming the quality of the proposed solutions.