On the Stability of Random Matrix Product with Markovian Noise: Application to Linear Stochastic Approximation and TD Learning - Archive ouverte HAL
Communication Dans Un Congrès Année : 2021

On the Stability of Random Matrix Product with Markovian Noise: Application to Linear Stochastic Approximation and TD Learning

Alexey Naumov
  • Fonction : Auteur
Sergey Samsonov
  • Fonction : Auteur
Hoi-To Wai
  • Fonction : Auteur

Résumé

This paper studies the exponential stability of random matrix products driven by a general (possibly unbounded) state space Markov chain. It is a cornerstone in the analysis of stochastic algorithms in machine learning (e.g. for parameter tracking in online learning or reinforcement learning). The existing results impose strong conditions such as uniform boundedness of the matrix-valued functions and uniform ergodicity of the Markov chains. Our main contribution is an exponential stability result for the $p$-th moment of random matrix product, provided that (i) the underlying Markov chain satisfies a super-Lyapunov drift condition, (ii) the growth of the matrix-valued functions is controlled by an appropriately defined function (related to the drift condition). Using this result, we give finite-time $p$-th moment bounds for constant and decreasing stepsize linear stochastic approximation schemes with Markovian noise on general state space. We illustrate these findings for linear value-function estimation in reinforcement learning. We provide finite-time $p$-th moment bound for various members of temporal difference (TD) family of algorithms.
Fichier principal
Vignette du fichier
2102.00185v1.pdf (475.58 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04398571 , version 1 (12-07-2024)

Identifiants

Citer

Alain Durmus, Eric Moulines, Alexey Naumov, Sergey Samsonov, Hoi-To Wai. On the Stability of Random Matrix Product with Markovian Noise: Application to Linear Stochastic Approximation and TD Learning. Conference on Learning Theory, Aug 2021, Boulder, United States. ⟨hal-04398571⟩
89 Consultations
21 Téléchargements

Altmetric

Partager

More