A simple dynamic bandit algorithm for hyper-parameter tuning

Xuedong Shang 1 Emilie Kaufmann 1 Michal Valko 2, 1
1 SEQUEL - Sequential Learning
Inria Lille - Nord Europe, CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189
2 DeepMind [Paris]
Abstract : Hyper-parameter tuning is a major part of modern machine learning systems. The tuning itself can be seen as a sequential resource allocation problem. As such, methods for multi-armed bandits have been already applied. In this paper, we view hyper-parameter optimization as an instance of best-arm identification in infinitely many-armed bandits. We propose D-TTTS, a new adaptive algorithm inspired by Thompson sampling, which dynamically balances between refining the estimate of the quality of hyper-parameter configurations previously explored and adding new hyper-parameter configurations to the pool of candidates. The algorithm is easy to implement and shows competitive performance compared to state-of-the-art algorithms for hyper-parameter tuning.
Document type :
Conference papers
Complete list of metadatas
https://hal.inria.fr/hal-02145200
Contributor : Michal Valko <>
Submitted on : Saturday, June 1, 2019 - 11:49:38 PM
Last modification on : Friday, June 14, 2019 - 11:39:58 AM

File

shang2019simple.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02145200, version 1

Collections

CRISTAL | INRIA | UNIV-LILLE

Citation

Xuedong Shang, Emilie Kaufmann, Michal Valko. A simple dynamic bandit algorithm for hyper-parameter tuning. Workshop on Automated Machine Learning at International Conference on Machine Learning, AutoML@ICML 2019 - 6th ICML Workshop on Automated Machine Learning, Jun 2019, Long Beach, United States. ⟨hal-02145200⟩

Export

BibTeX TEI DC DCterms EndNote

Share

Metrics

Record views

63

Files downloads

434

Contact


archive-ouverte@inria.fr