Gradient Estimation with Discrete Stein Operators

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Shi, Jiaxin, Zhou, Yuhao, Hwang, Jessica, Titsias, Michalis K., Mackey, Lester
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929312650756096
author Shi, Jiaxin
Zhou, Yuhao
Hwang, Jessica
Titsias, Michalis K.
Mackey, Lester
author_facet Shi, Jiaxin
Zhou, Yuhao
Hwang, Jessica
Titsias, Michalis K.
Mackey, Lester
contents Gradient estimation -- approximating the gradient of an expectation with respect to the parameters of a distribution -- is central to the solution of many machine learning problems. However, when the distribution is discrete, most common gradient estimators suffer from excessive variance. To improve the quality of gradient estimation, we introduce a variance reduction technique based on Stein operators for discrete distributions. We then use this technique to build flexible control variates for the REINFORCE leave-one-out estimator. Our control variates can be adapted online to minimize variance and do not require extra evaluations of the target function. In benchmark generative modeling tasks such as training binary variational autoencoders, our gradient estimator achieves substantially lower variance than state-of-the-art estimators with the same number of function evaluations.
format Preprint
id arxiv_https___arxiv_org_abs_2202_09497
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Gradient Estimation with Discrete Stein Operators
Shi, Jiaxin
Zhou, Yuhao
Hwang, Jessica
Titsias, Michalis K.
Mackey, Lester
Machine Learning
Gradient estimation -- approximating the gradient of an expectation with respect to the parameters of a distribution -- is central to the solution of many machine learning problems. However, when the distribution is discrete, most common gradient estimators suffer from excessive variance. To improve the quality of gradient estimation, we introduce a variance reduction technique based on Stein operators for discrete distributions. We then use this technique to build flexible control variates for the REINFORCE leave-one-out estimator. Our control variates can be adapted online to minimize variance and do not require extra evaluations of the target function. In benchmark generative modeling tasks such as training binary variational autoencoders, our gradient estimator achieves substantially lower variance than state-of-the-art estimators with the same number of function evaluations.
title Gradient Estimation with Discrete Stein Operators
topic Machine Learning
url https://arxiv.org/abs/2202.09497