Alpha-divergence minimization for deep Gaussian processes

This paper proposes the minimization of α-divergences for approximate inference in the context of deep Gaussian processes (DGPs). The proposed method can be considered as a generalization of variational inference (VI) and expectation propagation (EP), two previously used methods for approximate infe...

Descripción completa

Detalles Bibliográficos
Autores: Villacampa Calvo, Carlos, Hernández Muñoz, Gonzalo, Hernández Lobato, Daniel
Tipo de recurso: artículo
Fecha de publicación:2022
País:España
Institución:Universidad Autónoma de Madrid
Repositorio:Biblos-e Archivo. Repositorio Institucional de la UAM
Idioma:inglés
OAI Identifier:oai:repositorio.uam.es:10486/704028
Acceso en línea:http://hdl.handle.net/10486/704028
https://dx.doi.org/10.1016/j.ijar.2022.08.003
Access Level:acceso abierto
Palabra clave:Deep Gaussian processes
Expectation propagation
α-divergences
Approximate inference
Variational inference
Informática
id ES_303571ea061343da4dfbcd7faac8705e
oai_identifier_str oai:repositorio.uam.es:10486/704028
network_acronym_str ES
network_name_str España
repository_id_str
spelling Alpha-divergence minimization for deep Gaussian processesVillacampa Calvo, CarlosHernández Muñoz, GonzaloHernández Lobato, DanielDeep Gaussian processesExpectation propagationα-divergencesApproximate inferenceVariational inferenceInformáticaThis paper proposes the minimization of α-divergences for approximate inference in the context of deep Gaussian processes (DGPs). The proposed method can be considered as a generalization of variational inference (VI) and expectation propagation (EP), two previously used methods for approximate inference in DGPs. Both VI and EP are based on the minimization of the Kullback-Leibler divergence. The proposed method is based on a scalable version of power expectation propagation, a method that introduces an extra parameter α that specifies the targeted α-divergence to be optimized. In particular, such a method can recover the VI solution when α → 0 and the EP solution when α → 1. An exhaustive experimental evaluation shows that the minimization of α-divergences via the proposed method is feasible in DGPs and that choosing intermediate values of the α parameter between 0 and 1 can give better results in some problems. This means that one can improve the results of VI and EP when training DGPs. Importantly, the proposed method allows for stochastic optimization techniques, making it able to address datasets with several millions of instancesThe authors gratefully acknowledge the use of the facilities of Centro de Computación Científica (CCC) at Universidad Autónoma de Madrid. The authors also acknowledge financial support from Spanish Plan Nacional I+D+i, Ministerio de Ciencia e Innovación, grant PID2019-106827GB-I00 / AEI / 10.13039/501100011033ElsevierDepartamento de Ingeniería InformáticaEscuela Politécnica Superior20222022-08-22research articlehttp://purl.org/coar/resource_type/c_2df8fbb1VoRhttp://purl.org/coar/version/c_970fb48d4fbd8a85info:eu-repo/semantics/articleapplication/pdfhttp://hdl.handle.net/10486/704028https://dx.doi.org/10.1016/j.ijar.2022.08.003reponame:Biblos-e Archivo. Repositorio Institucional de la UAMinstname:Universidad Autónoma de MadridInglésengopen accesshttp://purl.org/coar/access_right/c_abf2info:eu-repo/semantics/openAccessoai:repositorio.uam.es:10486/7040282026-06-23T12:46:27Z
dc.title.none.fl_str_mv Alpha-divergence minimization for deep Gaussian processes
title Alpha-divergence minimization for deep Gaussian processes
spellingShingle Alpha-divergence minimization for deep Gaussian processes
Villacampa Calvo, Carlos
Deep Gaussian processes
Expectation propagation
α-divergences
Approximate inference
Variational inference
Informática
title_short Alpha-divergence minimization for deep Gaussian processes
title_full Alpha-divergence minimization for deep Gaussian processes
title_fullStr Alpha-divergence minimization for deep Gaussian processes
title_full_unstemmed Alpha-divergence minimization for deep Gaussian processes
title_sort Alpha-divergence minimization for deep Gaussian processes
dc.creator.none.fl_str_mv Villacampa Calvo, Carlos
Hernández Muñoz, Gonzalo
Hernández Lobato, Daniel
author Villacampa Calvo, Carlos
author_facet Villacampa Calvo, Carlos
Hernández Muñoz, Gonzalo
Hernández Lobato, Daniel
author_role author
author2 Hernández Muñoz, Gonzalo
Hernández Lobato, Daniel
author2_role author
author
dc.contributor.none.fl_str_mv Departamento de Ingeniería Informática
Escuela Politécnica Superior
dc.subject.none.fl_str_mv Deep Gaussian processes
Expectation propagation
α-divergences
Approximate inference
Variational inference
Informática
topic Deep Gaussian processes
Expectation propagation
α-divergences
Approximate inference
Variational inference
Informática
description This paper proposes the minimization of α-divergences for approximate inference in the context of deep Gaussian processes (DGPs). The proposed method can be considered as a generalization of variational inference (VI) and expectation propagation (EP), two previously used methods for approximate inference in DGPs. Both VI and EP are based on the minimization of the Kullback-Leibler divergence. The proposed method is based on a scalable version of power expectation propagation, a method that introduces an extra parameter α that specifies the targeted α-divergence to be optimized. In particular, such a method can recover the VI solution when α → 0 and the EP solution when α → 1. An exhaustive experimental evaluation shows that the minimization of α-divergences via the proposed method is feasible in DGPs and that choosing intermediate values of the α parameter between 0 and 1 can give better results in some problems. This means that one can improve the results of VI and EP when training DGPs. Importantly, the proposed method allows for stochastic optimization techniques, making it able to address datasets with several millions of instances
publishDate 2022
dc.date.none.fl_str_mv 2022
2022-08-22
dc.type.none.fl_str_mv research article
http://purl.org/coar/resource_type/c_2df8fbb1
VoR
http://purl.org/coar/version/c_970fb48d4fbd8a85
dc.type.openaire.fl_str_mv info:eu-repo/semantics/article
format article
dc.identifier.none.fl_str_mv http://hdl.handle.net/10486/704028
https://dx.doi.org/10.1016/j.ijar.2022.08.003
url http://hdl.handle.net/10486/704028
https://dx.doi.org/10.1016/j.ijar.2022.08.003
dc.language.none.fl_str_mv Inglés
eng
language_invalid_str_mv Inglés
language eng
dc.rights.none.fl_str_mv open access
http://purl.org/coar/access_right/c_abf2
dc.rights.openaire.fl_str_mv info:eu-repo/semantics/openAccess
rights_invalid_str_mv open access
http://purl.org/coar/access_right/c_abf2
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.publisher.none.fl_str_mv Elsevier
publisher.none.fl_str_mv Elsevier
dc.source.none.fl_str_mv reponame:Biblos-e Archivo. Repositorio Institucional de la UAM
instname:Universidad Autónoma de Madrid
instname_str Universidad Autónoma de Madrid
reponame_str Biblos-e Archivo. Repositorio Institucional de la UAM
collection Biblos-e Archivo. Repositorio Institucional de la UAM
repository.name.fl_str_mv
repository.mail.fl_str_mv
_version_ 1869405530086178816
score 15,300724