An Equivalence between Bayesian Priors and Penalties in Variational Inference

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wolinski, Pierre, Charpiat, Guillaume, Ollivier, Yann
Format: Preprint
Veröffentlicht: 2020
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913225287663616
author Wolinski, Pierre
Charpiat, Guillaume
Ollivier, Yann
author_facet Wolinski, Pierre
Charpiat, Guillaume
Ollivier, Yann
contents In machine learning, it is common to optimize the parameters of a probabilistic model, modulated by an ad hoc regularization term that penalizes some values of the parameters. Regularization terms appear naturally in Variational Inference, a tractable way to approximate Bayesian posteriors: the loss to optimize contains a Kullback--Leibler divergence term between the approximate posterior and a Bayesian prior. We fully characterize the regularizers that can arise according to this procedure, and provide a systematic way to compute the prior corresponding to a given penalty. Such a characterization can be used to discover constraints over the penalty function, so that the overall procedure remains Bayesian.
format Preprint
id arxiv_https___arxiv_org_abs_2002_00178
institution arXiv
publishDate 2020
record_format arxiv
spellingShingle An Equivalence between Bayesian Priors and Penalties in Variational Inference
Wolinski, Pierre
Charpiat, Guillaume
Ollivier, Yann
Machine Learning
Statistics Theory
In machine learning, it is common to optimize the parameters of a probabilistic model, modulated by an ad hoc regularization term that penalizes some values of the parameters. Regularization terms appear naturally in Variational Inference, a tractable way to approximate Bayesian posteriors: the loss to optimize contains a Kullback--Leibler divergence term between the approximate posterior and a Bayesian prior. We fully characterize the regularizers that can arise according to this procedure, and provide a systematic way to compute the prior corresponding to a given penalty. Such a characterization can be used to discover constraints over the penalty function, so that the overall procedure remains Bayesian.
title An Equivalence between Bayesian Priors and Penalties in Variational Inference
topic Machine Learning
Statistics Theory
url https://arxiv.org/abs/2002.00178