Neural Architecture Tuning: A BO-Powered NAS Tool

Ouertatani, Houssem; Maxim, Cristian; Niar, Smail; Talbi, El-Ghazali

Type de document :

Communication dans un congrès avec actes

Titre :

Neural Architecture Tuning: A BO-Powered NAS Tool

Auteur(s) :

Ouertatani, Houssem [Auteur]
IRT SystemX
Inria Lille - Nord Europe
Optimisation de grande taille et calcul large échelle [BONUS]
Maxim, Cristian [Auteur]
IRT SystemX
Niar, Smail [Auteur]
Université Polytechnique Hauts-de-France [UPHF]
Laboratoire d'Automatique, de Mécanique et d'Informatique industrielles et Humaines - UMR 8201 [LAMIH]
Talbi, El-Ghazali [Auteur]

Optimisation de grande taille et calcul large échelle [BONUS]
Université de Lille
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]

Titre de la manifestation scientifique :

International Conference in Optimization and Learning (OLA)

Ville :

Dubrovnik

Pays :

Croatie

Date de début de la manifestation scientifique :

2024-05-13

Mot(s)-clé(s) en anglais :

Neural Architecture Search
Bayesian Optimization
Custom search space
Multi-fidelity BO

Discipline(s) HAL :

Informatique [cs]

Résumé en anglais : [en]

Neural Architecture Search (NAS) consists of applying anoptimization technique to find the best performing architecture(s) in adefined search space, with regard to an objective function. The practicalimplementation of NAS ...
Lire la suite >Neural Architecture Search (NAS) consists of applying anoptimization technique to find the best performing architecture(s) in adefined search space, with regard to an objective function. The practicalimplementation of NAS currently carries certain limitations, includingprohibitive costs with the need for a large number of evaluations, an inflexibilityin defining the search space by often having to select from alimited set of possible design components, and a difficulty of integratingexisting architecture code by requiring a specialized design languagefor search space specification. We propose a simplified search tool, withefficiency in the number of evaluations needed to achieve good results,and flexibility by design, allowing for an easy and open definition of thesearch space and objective function. Interoperability with existing codeor newly released architectures from the literature allows the user toquickly and easily tune architectures to produce well-performing solutionstailor-made for particular use cases. We practically apply this toolto certain vision search spaces, and showcase its effectiveness.Lire moins >

Langue :

Anglais

Comité de lecture :

Oui

Audience :

Internationale

Vulgarisation :

Non

Collections :

Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189

Source :

Harvested from HAL

Fichiers

document
Accès libre
Accéder au document

ola_preprint.pdf
Accès libre
Accéder au document

Neural Architecture Tuning: A BO-Powered NAS Tool BibTeX CSV Excel RIS

Fichiers

Neural Architecture Tuning: A BO-Powered NAS Tool

BibTeX

CSV

Excel

RIS