Matches in SemOpenAlex for { <https://semopenalex.org/work/W4310996952> ?p ?o ?g. }
Showing items 1 to 89 of
89
with 100 items per page.
- W4310996952 abstract "Variational autoencoders (VAEs) are powerful tools for learning latent representations of data used in a wide range of applications. In practice, VAEs usually require multiple training rounds to choose the amount of information the latent variable should retain. This trade-off between the reconstruction error (distortion) and the KL divergence (rate) is typically parameterized by a hyperparameter $beta$. In this paper, we introduce Multi-Rate VAE (MR-VAE), a computationally efficient framework for learning optimal parameters corresponding to various $beta$ in a single training run. The key idea is to explicitly formulate a response function that maps $beta$ to the optimal parameters using hypernetworks. MR-VAEs construct a compact response hypernetwork where the pre-activations are conditionally gated based on $beta$. We justify the proposed architecture by analyzing linear VAEs and showing that it can represent response functions exactly for linear VAEs. With the learned hypernetwork, MR-VAEs can construct the rate-distortion curve without additional training and can be deployed with significantly less hyperparameter tuning. Empirically, our approach is competitive and often exceeds the performance of multiple $beta$-VAEs training with minimal computation and memory overheads." @default.
- W4310996952 created "2022-12-22" @default.
- W4310996952 creator A5004761297 @default.
- W4310996952 creator A5012276327 @default.
- W4310996952 creator A5019732625 @default.
- W4310996952 creator A5067036768 @default.
- W4310996952 creator A5072190856 @default.
- W4310996952 creator A5075190149 @default.
- W4310996952 creator A5078051293 @default.
- W4310996952 date "2022-12-07" @default.
- W4310996952 modified "2023-09-30" @default.
- W4310996952 title "Multi-Rate VAE: Train Once, Get the Full Rate-Distortion Curve" @default.
- W4310996952 doi "https://doi.org/10.48550/arxiv.2212.03905" @default.
- W4310996952 hasPublicationYear "2022" @default.
- W4310996952 type Work @default.
- W4310996952 citedByCount "0" @default.
- W4310996952 crossrefType "posted-content" @default.
- W4310996952 hasAuthorship W4310996952A5004761297 @default.
- W4310996952 hasAuthorship W4310996952A5012276327 @default.
- W4310996952 hasAuthorship W4310996952A5019732625 @default.
- W4310996952 hasAuthorship W4310996952A5067036768 @default.
- W4310996952 hasAuthorship W4310996952A5072190856 @default.
- W4310996952 hasAuthorship W4310996952A5075190149 @default.
- W4310996952 hasAuthorship W4310996952A5078051293 @default.
- W4310996952 hasBestOaLocation W43109969521 @default.
- W4310996952 hasConcept C11413529 @default.
- W4310996952 hasConcept C119857082 @default.
- W4310996952 hasConcept C126780896 @default.
- W4310996952 hasConcept C14036430 @default.
- W4310996952 hasConcept C153180895 @default.
- W4310996952 hasConcept C154945302 @default.
- W4310996952 hasConcept C159985019 @default.
- W4310996952 hasConcept C165464430 @default.
- W4310996952 hasConcept C17744445 @default.
- W4310996952 hasConcept C192562407 @default.
- W4310996952 hasConcept C194257627 @default.
- W4310996952 hasConcept C199360897 @default.
- W4310996952 hasConcept C199539241 @default.
- W4310996952 hasConcept C204323151 @default.
- W4310996952 hasConcept C2776257435 @default.
- W4310996952 hasConcept C2776359362 @default.
- W4310996952 hasConcept C2780801425 @default.
- W4310996952 hasConcept C31258907 @default.
- W4310996952 hasConcept C41008148 @default.
- W4310996952 hasConcept C45374587 @default.
- W4310996952 hasConcept C78458016 @default.
- W4310996952 hasConcept C8642999 @default.
- W4310996952 hasConcept C86803240 @default.
- W4310996952 hasConcept C94625758 @default.
- W4310996952 hasConceptScore W4310996952C11413529 @default.
- W4310996952 hasConceptScore W4310996952C119857082 @default.
- W4310996952 hasConceptScore W4310996952C126780896 @default.
- W4310996952 hasConceptScore W4310996952C14036430 @default.
- W4310996952 hasConceptScore W4310996952C153180895 @default.
- W4310996952 hasConceptScore W4310996952C154945302 @default.
- W4310996952 hasConceptScore W4310996952C159985019 @default.
- W4310996952 hasConceptScore W4310996952C165464430 @default.
- W4310996952 hasConceptScore W4310996952C17744445 @default.
- W4310996952 hasConceptScore W4310996952C192562407 @default.
- W4310996952 hasConceptScore W4310996952C194257627 @default.
- W4310996952 hasConceptScore W4310996952C199360897 @default.
- W4310996952 hasConceptScore W4310996952C199539241 @default.
- W4310996952 hasConceptScore W4310996952C204323151 @default.
- W4310996952 hasConceptScore W4310996952C2776257435 @default.
- W4310996952 hasConceptScore W4310996952C2776359362 @default.
- W4310996952 hasConceptScore W4310996952C2780801425 @default.
- W4310996952 hasConceptScore W4310996952C31258907 @default.
- W4310996952 hasConceptScore W4310996952C41008148 @default.
- W4310996952 hasConceptScore W4310996952C45374587 @default.
- W4310996952 hasConceptScore W4310996952C78458016 @default.
- W4310996952 hasConceptScore W4310996952C8642999 @default.
- W4310996952 hasConceptScore W4310996952C86803240 @default.
- W4310996952 hasConceptScore W4310996952C94625758 @default.
- W4310996952 hasLocation W43109969521 @default.
- W4310996952 hasOpenAccess W4310996952 @default.
- W4310996952 hasPrimaryLocation W43109969521 @default.
- W4310996952 hasRelatedWork W3014815208 @default.
- W4310996952 hasRelatedWork W3199608561 @default.
- W4310996952 hasRelatedWork W4210794429 @default.
- W4310996952 hasRelatedWork W4223456145 @default.
- W4310996952 hasRelatedWork W4283697347 @default.
- W4310996952 hasRelatedWork W4295309597 @default.
- W4310996952 hasRelatedWork W4295681619 @default.
- W4310996952 hasRelatedWork W4298144215 @default.
- W4310996952 hasRelatedWork W4309113015 @default.
- W4310996952 hasRelatedWork W4323894855 @default.
- W4310996952 isParatext "false" @default.
- W4310996952 isRetracted "false" @default.
- W4310996952 workType "article" @default.