Show simple item record

Radon Sobolev Variational Auto-Encoders

hal.structure.identifierCEntre de REcherches en MAthématiques de la DEcision [CEREMADE]
dc.contributor.authorTurinici, Gabriel
HAL ID: 16
ORCID: 0000-0003-2713-006X
dc.date.accessioned2021-10-28T11:11:28Z
dc.date.available2021-10-28T11:11:28Z
dc.date.issued2021
dc.identifier.issn0893-6080
dc.identifier.urihttps://basepub.dauphine.psl.eu/handle/123456789/22113
dc.language.isoenen
dc.subjectVariational Auto-Encoderen
dc.subjectGenerative modelen
dc.subjectSobolev spacesen
dc.subjectRadon Sobolev Variational Auto-Encoderen
dc.subject.ddc515en
dc.titleRadon Sobolev Variational Auto-Encodersen
dc.typeArticle accepté pour publication ou publié
dc.description.abstractenThe quality of generative models (such as Generative adversarial networks and Variational Auto-Encoders) depends heavily on the choice of a good probability distance. However some popular metrics like the Wasserstein or the Sliced Wasserstein distances, the Jensen–Shannon divergence, the Kullback–Leibler divergence, lack convenient properties such as (geodesic) convexity, fast evaluation and so on. To address these shortcomings, we introduce a class of distances that have built-in convexity. We investigate the relationship with some known paradigms (sliced distances – a synonym for Radon distances – reproducing kernel Hilbert spaces, energy distances). The distances are shown to possess fast implementations and are included in an adapted Variational Auto-Encoder termed Radon–Sobolev Variational Auto-Encoder (RS-VAE) which produces high quality results on standard generative datasets.en
dc.relation.isversionofjnlnameNeural Networks
dc.relation.isversionofjnlvol141en
dc.relation.isversionofjnldate2021-09
dc.relation.isversionofjnlpages294-305en
dc.relation.isversionofdoi10.1016/j.neunet.2021.04.018en
dc.relation.isversionofjnlpublisherElsevieren
dc.subject.ddclabelAnalyseen
dc.relation.forthcomingnonen
dc.description.ssrncandidatenon
dc.description.halcandidatenonen
dc.description.readershiprechercheen
dc.description.audienceInternationalen
dc.relation.Isversionofjnlpeerreviewedouien
dc.date.updated2021-10-28T09:46:11Z
hal.author.functionaut


Files in this item

Thumbnail
Name:
163541413353053.pdf
Size:
1.489Mb
Format:
PDF
View/Open

This item appears in the following Collection(s)

Show simple item record