Stochastic Optimization with Variance Reduction for Infinite Datasets with Finite-Sum Structure

Abstract : Stochastic optimization algorithms with variance reduction have proven successful for minimizing large finite sums of functions. Unfortunately, these techniques are unable to deal with stochastic perturbations of input data, induced for example by data augmentation. In such cases, the objective is no longer a finite sum, and the main candidate for optimization is the stochastic gradient descent method (SGD). In this paper, we introduce a variance reduction approach for these settings when the objective is strongly convex. After an initial linearly convergent phase, the algorithm achieves a $O(1/t)$ convergence rate in expectation like SGD, but with a constant factor that is typically much smaller, depending on the variance of gradient estimates due to perturbations on a single example. We also introduce extensions of the algorithm to composite objectives and non-uniform sampling.
Liste complète des métadonnées


https://hal.inria.fr/hal-01375816
Contributeur : Alberto Bietti <>
Soumis le : lundi 27 février 2017 - 14:21:42
Dernière modification le : jeudi 2 mars 2017 - 15:34:37

Fichiers

stoch-miso.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01375816, version 4
  • ARXIV : 1610.00970

Collections

Citation

Alberto Bietti, Julien Mairal. Stochastic Optimization with Variance Reduction for Infinite Datasets with Finite-Sum Structure. 2017. <hal-01375816v4>

Partager

Métriques

Consultations de
la notice

652

Téléchargements du document

69