Matches in SemOpenAlex for { <https://semopenalex.org/work/W4384647956> ?p ?o ?g. }
Showing items 1 to 73 of
73
with 100 items per page.
- W4384647956 abstract "In generative modeling, numerous successful approaches leverage a low-dimensional latent space, e.g., Stable Diffusion models the latent space induced by an encoder and generates images through a paired decoder. Although the selection of the latent space is empirically pivotal, determining the optimal choice and the process of identifying it remain unclear. In this study, we aim to shed light on this under-explored topic by rethinking the latent space from the perspective of model complexity. Our investigation starts with the classic generative adversarial networks (GANs). Inspired by the GAN training objective, we propose a novel distance between the latent and data distributions, whose minimization coincides with that of the generator complexity. The minimizer of this distance is characterized as the optimal data-dependent latent that most effectively capitalizes on the generator's capacity. Then, we consider parameterizing such a latent distribution by an encoder network and propose a two-stage training strategy called Decoupled Autoencoder (DAE), where the encoder is only updated in the first stage with an auxiliary decoder and then frozen in the second stage while the actual decoder is being trained. DAE can improve the latent distribution and as a result, improve the generative performance. Our theoretical analyses are corroborated by comprehensive experiments on various models such as VQGAN and Diffusion Transformer, where our modifications yield significant improvements in sample quality with decreased model complexity." @default.
- W4384647956 created "2023-07-19" @default.
- W4384647956 creator A5016338405 @default.
- W4384647956 creator A5027660820 @default.
- W4384647956 creator A5027853785 @default.
- W4384647956 creator A5036381691 @default.
- W4384647956 creator A5054709496 @default.
- W4384647956 creator A5077862962 @default.
- W4384647956 creator A5084382353 @default.
- W4384647956 date "2023-07-17" @default.
- W4384647956 modified "2023-09-25" @default.
- W4384647956 title "Complexity Matters: Rethinking the Latent Space for Generative Modeling" @default.
- W4384647956 doi "https://doi.org/10.48550/arxiv.2307.08283" @default.
- W4384647956 hasPublicationYear "2023" @default.
- W4384647956 type Work @default.
- W4384647956 citedByCount "0" @default.
- W4384647956 crossrefType "posted-content" @default.
- W4384647956 hasAuthorship W4384647956A5016338405 @default.
- W4384647956 hasAuthorship W4384647956A5027660820 @default.
- W4384647956 hasAuthorship W4384647956A5027853785 @default.
- W4384647956 hasAuthorship W4384647956A5036381691 @default.
- W4384647956 hasAuthorship W4384647956A5054709496 @default.
- W4384647956 hasAuthorship W4384647956A5077862962 @default.
- W4384647956 hasAuthorship W4384647956A5084382353 @default.
- W4384647956 hasBestOaLocation W43846479561 @default.
- W4384647956 hasConcept C101738243 @default.
- W4384647956 hasConcept C111919701 @default.
- W4384647956 hasConcept C11413529 @default.
- W4384647956 hasConcept C118505674 @default.
- W4384647956 hasConcept C119857082 @default.
- W4384647956 hasConcept C121332964 @default.
- W4384647956 hasConcept C153083717 @default.
- W4384647956 hasConcept C154945302 @default.
- W4384647956 hasConcept C163258240 @default.
- W4384647956 hasConcept C167966045 @default.
- W4384647956 hasConcept C2780992000 @default.
- W4384647956 hasConcept C39890363 @default.
- W4384647956 hasConcept C41008148 @default.
- W4384647956 hasConcept C50644808 @default.
- W4384647956 hasConcept C51167844 @default.
- W4384647956 hasConcept C62520636 @default.
- W4384647956 hasConceptScore W4384647956C101738243 @default.
- W4384647956 hasConceptScore W4384647956C111919701 @default.
- W4384647956 hasConceptScore W4384647956C11413529 @default.
- W4384647956 hasConceptScore W4384647956C118505674 @default.
- W4384647956 hasConceptScore W4384647956C119857082 @default.
- W4384647956 hasConceptScore W4384647956C121332964 @default.
- W4384647956 hasConceptScore W4384647956C153083717 @default.
- W4384647956 hasConceptScore W4384647956C154945302 @default.
- W4384647956 hasConceptScore W4384647956C163258240 @default.
- W4384647956 hasConceptScore W4384647956C167966045 @default.
- W4384647956 hasConceptScore W4384647956C2780992000 @default.
- W4384647956 hasConceptScore W4384647956C39890363 @default.
- W4384647956 hasConceptScore W4384647956C41008148 @default.
- W4384647956 hasConceptScore W4384647956C50644808 @default.
- W4384647956 hasConceptScore W4384647956C51167844 @default.
- W4384647956 hasConceptScore W4384647956C62520636 @default.
- W4384647956 hasLocation W43846479561 @default.
- W4384647956 hasOpenAccess W4384647956 @default.
- W4384647956 hasPrimaryLocation W43846479561 @default.
- W4384647956 hasRelatedWork W2769954154 @default.
- W4384647956 hasRelatedWork W2896475512 @default.
- W4384647956 hasRelatedWork W2903766720 @default.
- W4384647956 hasRelatedWork W2988134182 @default.
- W4384647956 hasRelatedWork W3201080032 @default.
- W4384647956 hasRelatedWork W3212002806 @default.
- W4384647956 hasRelatedWork W4286974158 @default.
- W4384647956 hasRelatedWork W4288024758 @default.
- W4384647956 hasRelatedWork W4299317306 @default.
- W4384647956 hasRelatedWork W4309969736 @default.
- W4384647956 isParatext "false" @default.
- W4384647956 isRetracted "false" @default.
- W4384647956 workType "article" @default.