Matches in SemOpenAlex for { <https://semopenalex.org/work/W3207330441> ?p ?o ?g. }
- W3207330441 abstract "Latent-space interpolation is commonly used to demonstrate the generalization ability of deep latent variable models. Various algorithms have been proposed to calculate the best trajectory between two encodings in the latent space. In this work, we show how data labeled with semantically continuous attributes can be utilized to conduct a quantitative evaluation of latent-space interpolation algorithms, for variational autoencoders. Our framework can be used to complement the standard qualitative comparison, and also enables evaluation for domains (such as graph) in which the visualization is difficult. Interestingly, our experiments reveal that the superiority of interpolation algorithms could be domain-dependent. While normalised interpolation works best for the image domain, spherical linear interpolation achieves the best performance in the graph domain. Next, we propose a simple-yet-effective method to restrict the latent space via a bottleneck structure in the encoder. We find that all interpolation algorithms evaluated in this work can benefit from this restriction. Finally, we conduct interpolation-aware training with the labeled attributes, and show that this explicit supervision can improve the interpolation performance." @default.
- W3207330441 created "2021-10-25" @default.
- W3207330441 creator A5026881155 @default.
- W3207330441 creator A5033632697 @default.
- W3207330441 creator A5037659256 @default.
- W3207330441 creator A5051747323 @default.
- W3207330441 creator A5056694985 @default.
- W3207330441 creator A5084194574 @default.
- W3207330441 date "2021-10-13" @default.
- W3207330441 modified "2023-10-11" @default.
- W3207330441 title "Revisiting Latent-Space Interpolation via a Quantitative Evaluation Framework." @default.
- W3207330441 cites W1834627138 @default.
- W3207330441 cites W1959608418 @default.
- W3207330441 cites W2009172320 @default.
- W3207330441 cites W2064076387 @default.
- W3207330441 cites W2099471712 @default.
- W3207330441 cites W2111708605 @default.
- W3207330441 cites W2141599568 @default.
- W3207330441 cites W2147774191 @default.
- W3207330441 cites W2163605009 @default.
- W3207330441 cites W2173520492 @default.
- W3207330441 cites W2184218725 @default.
- W3207330441 cites W2190691619 @default.
- W3207330441 cites W2522389179 @default.
- W3207330441 cites W2565330852 @default.
- W3207330441 cites W2610034660 @default.
- W3207330441 cites W2753738274 @default.
- W3207330441 cites W2806983170 @default.
- W3207330441 cites W2903764284 @default.
- W3207330441 cites W2906083776 @default.
- W3207330441 cites W2933036399 @default.
- W3207330441 cites W2944833253 @default.
- W3207330441 cites W2949900285 @default.
- W3207330441 cites W2952179887 @default.
- W3207330441 cites W2963104724 @default.
- W3207330441 cites W2963241486 @default.
- W3207330441 cites W2963366547 @default.
- W3207330441 cites W2963373786 @default.
- W3207330441 cites W2963733869 @default.
- W3207330441 cites W2964015378 @default.
- W3207330441 cites W2964121744 @default.
- W3207330441 cites W2964472976 @default.
- W3207330441 cites W2969280827 @default.
- W3207330441 cites W2970086547 @default.
- W3207330441 cites W2970485172 @default.
- W3207330441 cites W2998116985 @default.
- W3207330441 cites W2998313947 @default.
- W3207330441 cites W3011537551 @default.
- W3207330441 cites W3042719542 @default.
- W3207330441 cites W3109244019 @default.
- W3207330441 cites W3196466825 @default.
- W3207330441 hasPublicationYear "2021" @default.
- W3207330441 type Work @default.
- W3207330441 sameAs 3207330441 @default.
- W3207330441 citedByCount "0" @default.
- W3207330441 crossrefType "posted-content" @default.
- W3207330441 hasAuthorship W3207330441A5026881155 @default.
- W3207330441 hasAuthorship W3207330441A5033632697 @default.
- W3207330441 hasAuthorship W3207330441A5037659256 @default.
- W3207330441 hasAuthorship W3207330441A5051747323 @default.
- W3207330441 hasAuthorship W3207330441A5056694985 @default.
- W3207330441 hasAuthorship W3207330441A5084194574 @default.
- W3207330441 hasConcept C11413529 @default.
- W3207330441 hasConcept C115961682 @default.
- W3207330441 hasConcept C119768884 @default.
- W3207330441 hasConcept C134306372 @default.
- W3207330441 hasConcept C137800194 @default.
- W3207330441 hasConcept C143542097 @default.
- W3207330441 hasConcept C149635348 @default.
- W3207330441 hasConcept C153180895 @default.
- W3207330441 hasConcept C154945302 @default.
- W3207330441 hasConcept C171836373 @default.
- W3207330441 hasConcept C188325010 @default.
- W3207330441 hasConcept C203332170 @default.
- W3207330441 hasConcept C205203396 @default.
- W3207330441 hasConcept C207214200 @default.
- W3207330441 hasConcept C2780513914 @default.
- W3207330441 hasConcept C31972630 @default.
- W3207330441 hasConcept C33923547 @default.
- W3207330441 hasConcept C36503486 @default.
- W3207330441 hasConcept C41008148 @default.
- W3207330441 hasConcept C49608258 @default.
- W3207330441 hasConcept C51167844 @default.
- W3207330441 hasConceptScore W3207330441C11413529 @default.
- W3207330441 hasConceptScore W3207330441C115961682 @default.
- W3207330441 hasConceptScore W3207330441C119768884 @default.
- W3207330441 hasConceptScore W3207330441C134306372 @default.
- W3207330441 hasConceptScore W3207330441C137800194 @default.
- W3207330441 hasConceptScore W3207330441C143542097 @default.
- W3207330441 hasConceptScore W3207330441C149635348 @default.
- W3207330441 hasConceptScore W3207330441C153180895 @default.
- W3207330441 hasConceptScore W3207330441C154945302 @default.
- W3207330441 hasConceptScore W3207330441C171836373 @default.
- W3207330441 hasConceptScore W3207330441C188325010 @default.
- W3207330441 hasConceptScore W3207330441C203332170 @default.
- W3207330441 hasConceptScore W3207330441C205203396 @default.
- W3207330441 hasConceptScore W3207330441C207214200 @default.
- W3207330441 hasConceptScore W3207330441C2780513914 @default.
- W3207330441 hasConceptScore W3207330441C31972630 @default.
- W3207330441 hasConceptScore W3207330441C33923547 @default.