Critical Hyper-Parameters: No Random, No Cry

TL;DR

This paper advocates for the use of Low Discrepancy Sequences in hyperparameter optimization for deep learning, demonstrating they outperform random search in efficiency and effectiveness.

Contribution

It introduces a simple LDS-based method for hyperparameter search that is theoretically sound and practically effective, serving as a drop-in replacement for traditional methods.

Findings

LDS methods outperform random search in hyperparameter tuning efficiency.

LDS require fewer runs to find suitable hyperparameters in deep learning models.

The proposed LDS approach is versatile as a one-shot or initialization method.

Abstract

The selection of hyper-parameters is critical in Deep Learning. Because of the long training time of complex models and the availability of compute resources in the cloud, "one-shot" optimization schemes - where the sets of hyper-parameters are selected in advance (e.g. on a grid or in a random manner) and the training is executed in parallel - are commonly used. It is known that grid search is sub-optimal, especially when only a few critical parameters matter, and suggest to use random search instead. Yet, random search can be "unlucky" and produce sets of values that leave some part of the domain unexplored. Quasi-random methods, such as Low Discrepancy Sequences (LDS) avoid these issues. We show that such methods have theoretical properties that make them appealing for performing hyperparameter search, and demonstrate that, when applied to the selection of hyperparameters of complex Deep Learning models (such as state-of-the-art LSTM language models and image classification models), they yield suitable hyperparameters values with much fewer runs than random search. We propose a particularly simple LDS method which can be used as a drop-in replacement for grid or random search in any Deep Learning pipeline, both as a fully one-shot hyperparameter search or as an initializer in iterative batch optimization.