Passer à la navigation principale Passer à la recherche Passer au contenu principal

Tight High Probability Bounds for Linear Stochastic Approximation with Fixed Stepsize

  • ENS Paris-Saclay
  • National Research University
  • Université PSL
  • The Chinese University of Hong Kong

Résultats de recherche: Le chapitre dans un livre, un rapport, une anthologie ou une collectionContribution à une conférenceRevue par des pairs

Résumé

This paper provides a non-asymptotic analysis of linear stochastic approximation (LSA) algorithms with fixed stepsize. This family of methods arises in many machine learning tasks and is used to obtain approximate solutions of a linear system Āθ =¯b for which Ā and¯b can only be accessed through random estimates {(An, bn): n ∈ ℕ*}. Our analysis is based on new results regarding moments and high probability bounds for products of matrices which are shown to be tight. We derive high probability bounds on the performance of LSA under weaker conditions on the sequence {(An, bn): n ∈ ℕ*} than previous works. However, in contrast, we establish polynomial concentration bounds with order depending on the stepsize. We show that our conclusions cannot be improved without additional assumptions on the sequence of random matrices {An : n ∈ ℕ*}, and in particular that no Gaussian or exponential high probability bounds can hold. Finally, we pay a particular attention to establishing bounds with sharp order with respect to the number of iterations and the stepsize and whose leading terms contain the covariance matrices appearing in the central limit theorems.

langue originaleAnglais
titreAdvances in Neural Information Processing Systems 34 - 35th Conference on Neural Information Processing Systems, NeurIPS 2021
rédacteurs en chefMarc'Aurelio Ranzato, Alina Beygelzimer, Yann Dauphin, Percy S. Liang, Jenn Wortman Vaughan
EditeurNeural information processing systems foundation
Pages30063-30074
Nombre de pages12
ISBN (Electronique)9781713845393
étatPublié - 1 janv. 2021
Modification externeOui
Evénement35th Conference on Neural Information Processing Systems, NeurIPS 2021 - Virtual, Online
Durée: 6 déc. 202114 déc. 2021

Série de publications

NomAdvances in Neural Information Processing Systems
Volume36
ISSN (imprimé)1049-5258

Une conférence

Une conférence35th Conference on Neural Information Processing Systems, NeurIPS 2021
La villeVirtual, Online
période6/12/2114/12/21

Empreinte digitale

Examiner les sujets de recherche de « Tight High Probability Bounds for Linear Stochastic Approximation with Fixed Stepsize ». Ensemble, ils forment une empreinte digitale unique.

Contient cette citation