Optimal Computational Trade-Off of Inexact Proximal Methods (short version) - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2012

Optimal Computational Trade-Off of Inexact Proximal Methods (short version)

Résumé

In this paper, we investigate the trade-off between convergence rate and computational cost when minimizing a composite functional with proximal-gradient methods, which are popular optimisation tools in machine learning. We consider the case when the proximity operator is approximated via an iterative procedure, which yields algorithms with two nested loops. We show that the strategy minimizing the computational cost to reach a desired accuracy in finite time is to keep the number of inner iterations constant, which differs from the strategy indicated by a convergence rate analysis.
Fichier principal
Vignette du fichier
proxTradeoffWS.pdf (76.28 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-00771722 , version 1 (09-01-2013)

Identifiants

  • HAL Id : hal-00771722 , version 1

Citer

Pierre Machart, Luca Baldassarre, Sandrine Anthoine. Optimal Computational Trade-Off of Inexact Proximal Methods (short version). Multi-Trade-offs in Machine Learning (NIPS), Dec 2012, Lake Tahoe, United States. ⟨hal-00771722⟩
524 Consultations
156 Téléchargements

Partager

Gmail Facebook X LinkedIn More