Toward Optimal Run Racing: Application to Deep Learning Calibration - Inria - Institut national de recherche en sciences et technologies du numérique
Pré-Publication, Document De Travail Année : 2017

Toward Optimal Run Racing: Application to Deep Learning Calibration

Olivier Bousquet
Sylvain Gelly
  • Fonction : Auteur
Karol Kurach
  • Fonction : Auteur
Olivier Teytaud
Damien Vincent
  • Fonction : Auteur

Résumé

This paper aims at one-shot learning of deep neural nets, where a highly parallel setting is considered to address the algorithm calibration problem - selecting the best neural architecture and learning hyper-parameter values depending on the dataset at hand. The notoriously expensive calibration problem is optimally reduced by detecting and early stopping non-optimal runs. The theoretical contribution regards the optimality guarantees within the multiple hypothesis testing framework. Experimentations on the Cifar10, PTB and Wiki benchmarks demonstrate the relevance of the approach with a principled and consistent improvement on the state of the art with no extra hyper-parameter.

Dates et versions

hal-01634381 , version 1 (14-11-2017)

Identifiants

Citer

Olivier Bousquet, Sylvain Gelly, Karol Kurach, Marc Schoenauer, Michèle Sebag, et al.. Toward Optimal Run Racing: Application to Deep Learning Calibration. 2017. ⟨hal-01634381⟩
158 Consultations
0 Téléchargements

Altmetric

Partager

More