Simple view
Full metadata view
Authors
Statistics
Cramer-Wold Auto-Encoder
The computation of the distance to the true distribution is a key component of most state-of-the-art generative models. Inspired by prior works on the Sliced-Wasserstein Auto-Encoders(SWAE) and the Wasserstein Auto-Encoders with MMD-based penalty (WAE-MMD), wepropose a new generative model – a Cramer-Wold Auto-Encoder (CWAE). A fundamentalcomponent of CWAE is the characteristic kernel, the construction of which is one of the goalsof this paper, from here on referred to as the Cramer-Wold kernel. Its main distinguishingfeature is that it has a closed-form of the kernel product of radial Gaussians. Consequently,CWAE model has a closed-form for the distance between the posterior and the normal prior,which simplifies the optimization procedure by removing the need to sample in order tocompute the loss function. At the same time, CWAE performance often improves uponWAE-MMD and SWAE on standard benchmarks.Keywords:Auto-Encoder, Generative model, Wasserstein Auto-Encoder, Cramer-WoldTheorem, Deep neural network
dc.abstract.en | The computation of the distance to the true distribution is a key component of most state-of-the-art generative models. Inspired by prior works on the Sliced-Wasserstein Auto-Encoders(SWAE) and the Wasserstein Auto-Encoders with MMD-based penalty (WAE-MMD), wepropose a new generative model – a Cramer-Wold Auto-Encoder (CWAE). A fundamentalcomponent of CWAE is the characteristic kernel, the construction of which is one of the goalsof this paper, from here on referred to as the Cramer-Wold kernel. Its main distinguishingfeature is that it has a closed-form of the kernel product of radial Gaussians. Consequently,CWAE model has a closed-form for the distance between the posterior and the normal prior,which simplifies the optimization procedure by removing the need to sample in order tocompute the loss function. At the same time, CWAE performance often improves uponWAE-MMD and SWAE on standard benchmarks.Keywords:Auto-Encoder, Generative model, Wasserstein Auto-Encoder, Cramer-WoldTheorem, Deep neural network | pl |
dc.affiliation | Wydział Matematyki i Informatyki | pl |
dc.affiliation | Wydział Matematyki i Informatyki : Instytut Informatyki Analitycznej | pl |
dc.affiliation | Wydział Matematyki i Informatyki : Instytut Informatyki i Matematyki Komputerowej | pl |
dc.contributor.author | Knop, Szymon - 177158 | pl |
dc.contributor.author | Spurek, Przemysław - 135993 | pl |
dc.contributor.author | Tabor, Jacek - 132362 | pl |
dc.contributor.author | Podolak, Igor - 100165 | pl |
dc.contributor.author | Mazur, Marcin - 130444 | pl |
dc.contributor.author | Jastrzębski, Stanisław - 207335 | pl |
dc.date.accession | 2020-10-27 | pl |
dc.date.accessioned | 2020-10-29T19:05:49Z | |
dc.date.available | 2020-10-29T19:05:49Z | |
dc.date.issued | 2020 | pl |
dc.date.openaccess | 0 | |
dc.description.accesstime | w momencie opublikowania | |
dc.description.physical | 1-28 | pl |
dc.description.version | ostateczna wersja wydawcy | |
dc.description.volume | 21 | pl |
dc.identifier.eissn | 1533-7928 | pl |
dc.identifier.issn | 1532-4435 | pl |
dc.identifier.project | 2019/33/B/ST6/00894 | pl |
dc.identifier.project | 2017/25/B/ST6/01271 | pl |
dc.identifier.project | POIR.04.04.00-00-14DE/18-00 | pl |
dc.identifier.project | ROD UJ / OP | pl |
dc.identifier.uri | https://ruj.uj.edu.pl/xmlui/handle/item/251876 | |
dc.identifier.weblink | http://jmlr.org/papers/v21/19-560.html | pl |
dc.language | eng | pl |
dc.language.container | eng | pl |
dc.rights | Udzielam licencji. Uznanie autorstwa 4.0 Międzynarodowa | * |
dc.rights.licence | CC-BY | |
dc.rights.uri | http://creativecommons.org/licenses/by/4.0/legalcode.pl | * |
dc.share.type | otwarte czasopismo | |
dc.subtype | Article | pl |
dc.title | Cramer-Wold Auto-Encoder | pl |
dc.title.journal | Journal of Machine Learning Research | pl |
dc.type | JournalArticle | pl |
dspace.entity.type | Publication |