Hyperparameter Optimization via Sequential Uniform Designs

Published: 17 May 2023, Last Modified: 17 Sept 2023AutoML-Conf 2022 (Journal Track)Readers: Everyone
Link To Paper: https://jmlr.org/papers/v22/20-058.html
Journal Of Paper: Journal of Machine Learning Research
Confirmed Open Access: Yes
Topics From Call For Papers: Hyperparameter Optimization (HPO)
Broader Impact Statement On Ethical And Societal Implications: Machine learning and artificial intelligence are everywhere today, impacting virtually every industrial sector. The machine learning models are becoming increasingly complex and the number of hyperparameters explodes, so we usually need spend a huge amount of time and energy on parameter tuning or automated machine learning (AutoML). This raise concerns around the growing carbon footprint as it may potentially contribute to climate change. This paper proposes a novel efficient hyperparameter optimization (HPO) method based on the sequential uniform design (SeqUD) on the hyperparameter space. It is justified by the paper that SeqUD can be a promising and competitive alternative to existing AutML tools. It may reduce the computing time and help mitigate the concern around the growing carbon footprint of machine learning.
Reproducibility Checklist: pdf
Community Implementations: [![CatalyzeX](/images/catalyzex_icon.svg) 3 code implementations](https://www.catalyzex.com/paper/arxiv:2009.03586/code)
0 Replies