Fourier Stochastic Backpropagation - Archive ouverte HAL
Pré-Publication, Document De Travail Année : 2020

Fourier Stochastic Backpropagation

Joachim Flocon Cholet
  • Fonction : Auteur
  • PersonId : 1079369
Stéphane Gosselin
Sandrine Vaton

Résumé

Backpropagating gradients through random variables is at the heart of numerous machine learning applications. In this paper, we present a general framework for deriving stochastic backpropagation rules for any distribution, discrete or continuous. Our approach exploits the link between the characteristic function and the Fourier transform, to transport the derivatives from the parameters of the distribution to the random variable. Our method generalizes previously known estimators, and results in new estimators for the gamma, beta, Dirichlet and Laplace distributions. Furthermore, we show that the classical deterministic backproapagation rule and the discrete random variable case, can also be interpreted through stochastic backpropagation.
Fichier principal
Vignette du fichier
Fourier_Stochastic_Backpropagation.pdf (10.45 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-02968975 , version 1 (16-10-2020)
hal-02968975 , version 2 (05-12-2020)
hal-02968975 , version 3 (13-01-2021)

Identifiants

  • HAL Id : hal-02968975 , version 2

Citer

Amine Echraibi, Joachim Flocon Cholet, Stéphane Gosselin, Sandrine Vaton. Fourier Stochastic Backpropagation. 2020. ⟨hal-02968975v2⟩
388 Consultations
347 Téléchargements

Partager

More