Show simple item record

dc.contributor.authorRomero Merino, Enrique
dc.contributor.authorMazzanti Castrillejo, Fernando Pablo
dc.contributor.authorDelgado Pin, Jordi
dc.contributor.otherUniversitat Politècnica de Catalunya. Departament de Ciències de la Computació
dc.contributor.otherUniversitat Politècnica de Catalunya. Departament de Física
dc.date.accessioned2018-12-13T08:22:27Z
dc.date.available2018-12-13T08:22:27Z
dc.date.issued2018-07
dc.identifier.citationRomero, E., Mazzanti, F., Delgado, J. Neighborhood-based stopping criterion for contrastive divergence. "IEEE transactions on neural networks and learning systems", Juliol 2018, vol. 29, núm. 7, p. 2695-2704.
dc.identifier.issn2162-237X
dc.identifier.urihttp://hdl.handle.net/2117/125738
dc.description.abstractRestricted Boltzmann Machines (RBMs) are general unsupervised learning devices to ascertain generative models of data distributions. RBMs are often trained using the Contrastive Divergence (CD) learning algorithm, an approximation to the gradient of the data log-likelihood (logL). A simple reconstruction error is often used as a stopping criterion for CD, although several authors have raised doubts concerning the feasibility of this procedure. In many cases, the evolution curve of the reconstruction error is monotonic, while the logL is not, thus indicating that the former is not a good estimator of the optimal stopping point for learning. However, not many alternatives to the reconstruction error have been discussed in the literature. An estimation of the logL of the training data based on annealed importance sampling is feasible but computationally very expensive. In this manuscript, we present a simple and cheap alternative, based on the inclusion of information contained in neighboring states to the training set, as a stopping criterion for CD learning.
dc.format.extent10 p.
dc.language.isoeng
dc.subjectÀrees temàtiques de la UPC::Informàtica::Intel·ligència artificial::Aprenentatge automàtic
dc.subject.lcshMachine learning
dc.subject.lcshNeural networks (Computer science)
dc.subject.otherRecurrent neural networks
dc.subject.otherRestricted Boltzmann machines
dc.subject.otherUnsupervised learning
dc.titleNeighborhood-based stopping criterion for contrastive divergence
dc.typeArticle
dc.subject.lemacAprenentatge automàtic
dc.subject.lemacXarxes neuronals (Informàtica)
dc.contributor.groupUniversitat Politècnica de Catalunya. SIMCON - First-principles approaches to condensed matter physics: quantum effects and complexity
dc.contributor.groupUniversitat Politècnica de Catalunya. LARCA - Laboratori d'Algorísmia Relacional, Complexitat i Aprenentatge
dc.identifier.doi10.1109/TNNLS.2017.2697455
dc.description.peerreviewedPeer Reviewed
dc.relation.publisherversionhttps://ieeexplore.ieee.org/document/7930408
dc.rights.accessOpen Access
local.identifier.drac23537879
dc.description.versionPostprint (author's final draft)
dc.relation.projectidinfo:eu-repo/grantAgreement/MINECO/1PE/TIN2016-79576-R
dc.relation.projectidinfo:eu-repo/grantAgreement/MINECO//FIS2014-56257-C2-1-P/ES/MATERIA CUÁNTICA ULTRAFRIA/
dc.relation.projectidinfo:eu-repo/grantAgreement/AGAUR/V PRI/2014 SGR 890
dc.relation.projectidinfo:eu-repo/grantAgreement/MINECO//TIN2014-57226-P/ES/APRENDIZAJE COMPUTACIONAL Y COMUNICACION/
local.citation.authorRomero, E.; Mazzanti, F.; Delgado, J.
local.citation.publicationNameIEEE transactions on neural networks and learning systems
local.citation.volume29
local.citation.number7
local.citation.startingPage2695
local.citation.endingPage2704


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record