Matches in SemOpenAlex for { <https://semopenalex.org/work/W4289104496> ?p ?o ?g. }
Showing items 1 to 77 of
77
with 100 items per page.
- W4289104496 abstract "An implicit goal in works on deep generative models is that such models should be able to generate novel examples that were not previously seen in the training data. In this paper, we investigate to what extent this property holds for widely employed variational autoencoder (VAE) architectures. VAEs maximize a lower bound on the log marginal likelihood, which implies that they will in principle overfit the training data when provided with a sufficiently expressive decoder. In the limit of an infinite capacity decoder, the optimal generative model is a uniform mixture over the training data. More generally, an optimal decoder should output a weighted average over the examples in the training data, where the magnitude of the weights is determined by the proximity in the latent space. This leads to the hypothesis that, for a sufficiently high capacity encoder and decoder, the VAE decoder will perform nearest-neighbor matching according to the coordinates in the latent space. To test this hypothesis, we investigate generalization on the MNIST dataset. We consider both generalization to new examples of previously seen classes, and generalization to the classes that were withheld from the training set. In both cases, we find that reconstructions are closely approximated by nearest neighbors for higher-dimensional parameterizations. When generalizing to unseen classes however, lower-dimensional parameterizations offer a clear advantage." @default.
- W4289104496 created "2022-08-01" @default.
- W4289104496 creator A5042038501 @default.
- W4289104496 creator A5053808736 @default.
- W4289104496 creator A5059517243 @default.
- W4289104496 creator A5073129092 @default.
- W4289104496 creator A5079980080 @default.
- W4289104496 date "2018-12-22" @default.
- W4289104496 modified "2023-09-26" @default.
- W4289104496 title "Can VAEs Generate Novel Examples?" @default.
- W4289104496 doi "https://doi.org/10.48550/arxiv.1812.09624" @default.
- W4289104496 hasPublicationYear "2018" @default.
- W4289104496 type Work @default.
- W4289104496 citedByCount "0" @default.
- W4289104496 crossrefType "posted-content" @default.
- W4289104496 hasAuthorship W4289104496A5042038501 @default.
- W4289104496 hasAuthorship W4289104496A5053808736 @default.
- W4289104496 hasAuthorship W4289104496A5059517243 @default.
- W4289104496 hasAuthorship W4289104496A5073129092 @default.
- W4289104496 hasAuthorship W4289104496A5079980080 @default.
- W4289104496 hasBestOaLocation W42891044961 @default.
- W4289104496 hasConcept C101738243 @default.
- W4289104496 hasConcept C105795698 @default.
- W4289104496 hasConcept C108583219 @default.
- W4289104496 hasConcept C111919701 @default.
- W4289104496 hasConcept C11413529 @default.
- W4289104496 hasConcept C134306372 @default.
- W4289104496 hasConcept C153180895 @default.
- W4289104496 hasConcept C154945302 @default.
- W4289104496 hasConcept C165064840 @default.
- W4289104496 hasConcept C167966045 @default.
- W4289104496 hasConcept C177148314 @default.
- W4289104496 hasConcept C177264268 @default.
- W4289104496 hasConcept C190502265 @default.
- W4289104496 hasConcept C199360897 @default.
- W4289104496 hasConcept C22019652 @default.
- W4289104496 hasConcept C2778572836 @default.
- W4289104496 hasConcept C33923547 @default.
- W4289104496 hasConcept C39890363 @default.
- W4289104496 hasConcept C41008148 @default.
- W4289104496 hasConcept C50644808 @default.
- W4289104496 hasConceptScore W4289104496C101738243 @default.
- W4289104496 hasConceptScore W4289104496C105795698 @default.
- W4289104496 hasConceptScore W4289104496C108583219 @default.
- W4289104496 hasConceptScore W4289104496C111919701 @default.
- W4289104496 hasConceptScore W4289104496C11413529 @default.
- W4289104496 hasConceptScore W4289104496C134306372 @default.
- W4289104496 hasConceptScore W4289104496C153180895 @default.
- W4289104496 hasConceptScore W4289104496C154945302 @default.
- W4289104496 hasConceptScore W4289104496C165064840 @default.
- W4289104496 hasConceptScore W4289104496C167966045 @default.
- W4289104496 hasConceptScore W4289104496C177148314 @default.
- W4289104496 hasConceptScore W4289104496C177264268 @default.
- W4289104496 hasConceptScore W4289104496C190502265 @default.
- W4289104496 hasConceptScore W4289104496C199360897 @default.
- W4289104496 hasConceptScore W4289104496C22019652 @default.
- W4289104496 hasConceptScore W4289104496C2778572836 @default.
- W4289104496 hasConceptScore W4289104496C33923547 @default.
- W4289104496 hasConceptScore W4289104496C39890363 @default.
- W4289104496 hasConceptScore W4289104496C41008148 @default.
- W4289104496 hasConceptScore W4289104496C50644808 @default.
- W4289104496 hasLocation W42891044961 @default.
- W4289104496 hasOpenAccess W4289104496 @default.
- W4289104496 hasPrimaryLocation W42891044961 @default.
- W4289104496 hasRelatedWork W2806319940 @default.
- W4289104496 hasRelatedWork W2851950156 @default.
- W4289104496 hasRelatedWork W2885904013 @default.
- W4289104496 hasRelatedWork W2891962740 @default.
- W4289104496 hasRelatedWork W2906547788 @default.
- W4289104496 hasRelatedWork W2924440852 @default.
- W4289104496 hasRelatedWork W3023776785 @default.
- W4289104496 hasRelatedWork W3160820590 @default.
- W4289104496 hasRelatedWork W4221153573 @default.
- W4289104496 hasRelatedWork W4289104496 @default.
- W4289104496 isParatext "false" @default.
- W4289104496 isRetracted "false" @default.
- W4289104496 workType "article" @default.