A Stochastic Path-Integrated Differential EstimatoR Expectation Maximization Algorithm - Centre de mathématiques appliquées (CMAP) Accéder directement au contenu
Communication Dans Un Congrès Année : 2020

A Stochastic Path-Integrated Differential EstimatoR Expectation Maximization Algorithm

Résumé

The Expectation Maximization (EM) algorithm is of key importance for inference in latent variable models including mixture of regressors and experts, missing observations. This paper introduces a novel EM algorithm, called SPIDER-EM, for inference from a training set of size n, n ≫ 1. At the core of our algorithm is an estimator of the full conditional expectation in the E-step, adapted from the stochastic path-integrated differential estimator (SPIDER) technique. We derive finite-time complexity bounds for smooth non-convex likelihood: we show that for convergence to an ǫ-approximate stationary point, the complexity scales as K Opt (n, ǫ) = O(ǫ −1) and K CE (n, ǫ) = n + √ nO(ǫ −1), where K Opt (n, ǫ) and K CE (n, ǫ) are respectively the number of M-steps and the number of per-sample conditional expectations evaluations. This improves over the state-of-the-art algorithms. Numerical results support our findings.
Fichier principal
Vignette du fichier
main_final_HAL.pdf (696.04 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03029700 , version 1 (28-11-2020)

Identifiants

  • HAL Id : hal-03029700 , version 1

Citer

Gersende Fort, Eric Moulines, Hoi-To Wai. A Stochastic Path-Integrated Differential EstimatoR Expectation Maximization Algorithm. NeurIPS 2020 - 34th Conference on Neural Information Processing Systems, Dec 2020, Vancouver / Virtuel, Canada. ⟨hal-03029700⟩
144 Consultations
58 Téléchargements

Partager

Gmail Facebook X LinkedIn More