Salvato in:
Dettagli Bibliografici
Autori principali: Alon, Itai, Globerson, Amir, Wiesel, Ami
Natura: Preprint
Pubblicazione: 2021
Soggetti:
Accesso online:https://arxiv.org/abs/2110.13452
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913340356296704
author Alon, Itai
Globerson, Amir
Wiesel, Ami
author_facet Alon, Itai
Globerson, Amir
Wiesel, Ami
contents Generative models have been successfully used for generating realistic signals. Because the likelihood function is typically intractable in most of these models, the common practice is to use "implicit" models that avoid likelihood calculation. However, it is hard to obtain theoretical guarantees for such models. In particular, it is not understood when they can globally optimize their non-convex objectives. Here we provide such an analysis for the case of Maximum Mean Discrepancy (MMD) learning of generative models. We prove several optimality results, including for a Gaussian distribution with low rank covariance (where likelihood is inapplicable) and a mixture of Gaussians. Our analysis shows that that the MMD optimization landscape is benign in these cases, and therefore gradient based methods will globally minimize the MMD objective.
format Preprint
id arxiv_https___arxiv_org_abs_2110_13452
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle On the Optimization Landscape of Maximum Mean Discrepancy
Alon, Itai
Globerson, Amir
Wiesel, Ami
Machine Learning
Generative models have been successfully used for generating realistic signals. Because the likelihood function is typically intractable in most of these models, the common practice is to use "implicit" models that avoid likelihood calculation. However, it is hard to obtain theoretical guarantees for such models. In particular, it is not understood when they can globally optimize their non-convex objectives. Here we provide such an analysis for the case of Maximum Mean Discrepancy (MMD) learning of generative models. We prove several optimality results, including for a Gaussian distribution with low rank covariance (where likelihood is inapplicable) and a mixture of Gaussians. Our analysis shows that that the MMD optimization landscape is benign in these cases, and therefore gradient based methods will globally minimize the MMD objective.
title On the Optimization Landscape of Maximum Mean Discrepancy
topic Machine Learning
url https://arxiv.org/abs/2110.13452