Matches in SemOpenAlex for { <https://semopenalex.org/work/W4366999746> ?p ?o ?g. }
Showing items 1 to 81 of
81
with 100 items per page.
- W4366999746 abstract "Deep equilibrium models (DEQs) have proven to be very powerful for learning data representations. The idea is to replace traditional (explicit) feedforward neural networks with an implicit fixed-point equation, which allows to decouple the forward and backward passes. In particular, training DEQ layers becomes very memory-efficient via the implicit function theorem. However, backpropagation through DEQ layers still requires solving an expensive Jacobian-based equation. In this paper, we introduce a simple but effective strategy to avoid this computational burden. Our method relies on the Jacobian approximation of Broyden's method after the forward pass to compute the gradients during the backward pass. Experiments show that simply re-using this approximation can significantly speed up the training while not causing any performance degradation." @default.
- W4366999746 created "2023-04-27" @default.
- W4366999746 creator A5044769639 @default.
- W4366999746 creator A5048602838 @default.
- W4366999746 date "2023-04-23" @default.
- W4366999746 modified "2023-09-25" @default.
- W4366999746 title "Efficient Training of Deep Equilibrium Models" @default.
- W4366999746 doi "https://doi.org/10.48550/arxiv.2304.11663" @default.
- W4366999746 hasPublicationYear "2023" @default.
- W4366999746 type Work @default.
- W4366999746 citedByCount "0" @default.
- W4366999746 crossrefType "posted-content" @default.
- W4366999746 hasAuthorship W4366999746A5044769639 @default.
- W4366999746 hasAuthorship W4366999746A5048602838 @default.
- W4366999746 hasBestOaLocation W43669997461 @default.
- W4366999746 hasConcept C108583219 @default.
- W4366999746 hasConcept C111472728 @default.
- W4366999746 hasConcept C11413529 @default.
- W4366999746 hasConcept C121332964 @default.
- W4366999746 hasConcept C126255220 @default.
- W4366999746 hasConcept C127413603 @default.
- W4366999746 hasConcept C133731056 @default.
- W4366999746 hasConcept C138885662 @default.
- W4366999746 hasConcept C14036430 @default.
- W4366999746 hasConcept C153294291 @default.
- W4366999746 hasConcept C154945302 @default.
- W4366999746 hasConcept C155032097 @default.
- W4366999746 hasConcept C200331156 @default.
- W4366999746 hasConcept C2524010 @default.
- W4366999746 hasConcept C2777211547 @default.
- W4366999746 hasConcept C2780586882 @default.
- W4366999746 hasConcept C28719098 @default.
- W4366999746 hasConcept C28826006 @default.
- W4366999746 hasConcept C33923547 @default.
- W4366999746 hasConcept C38858127 @default.
- W4366999746 hasConcept C41008148 @default.
- W4366999746 hasConcept C47702885 @default.
- W4366999746 hasConcept C50644808 @default.
- W4366999746 hasConcept C78458016 @default.
- W4366999746 hasConcept C86803240 @default.
- W4366999746 hasConceptScore W4366999746C108583219 @default.
- W4366999746 hasConceptScore W4366999746C111472728 @default.
- W4366999746 hasConceptScore W4366999746C11413529 @default.
- W4366999746 hasConceptScore W4366999746C121332964 @default.
- W4366999746 hasConceptScore W4366999746C126255220 @default.
- W4366999746 hasConceptScore W4366999746C127413603 @default.
- W4366999746 hasConceptScore W4366999746C133731056 @default.
- W4366999746 hasConceptScore W4366999746C138885662 @default.
- W4366999746 hasConceptScore W4366999746C14036430 @default.
- W4366999746 hasConceptScore W4366999746C153294291 @default.
- W4366999746 hasConceptScore W4366999746C154945302 @default.
- W4366999746 hasConceptScore W4366999746C155032097 @default.
- W4366999746 hasConceptScore W4366999746C200331156 @default.
- W4366999746 hasConceptScore W4366999746C2524010 @default.
- W4366999746 hasConceptScore W4366999746C2777211547 @default.
- W4366999746 hasConceptScore W4366999746C2780586882 @default.
- W4366999746 hasConceptScore W4366999746C28719098 @default.
- W4366999746 hasConceptScore W4366999746C28826006 @default.
- W4366999746 hasConceptScore W4366999746C33923547 @default.
- W4366999746 hasConceptScore W4366999746C38858127 @default.
- W4366999746 hasConceptScore W4366999746C41008148 @default.
- W4366999746 hasConceptScore W4366999746C47702885 @default.
- W4366999746 hasConceptScore W4366999746C50644808 @default.
- W4366999746 hasConceptScore W4366999746C78458016 @default.
- W4366999746 hasConceptScore W4366999746C86803240 @default.
- W4366999746 hasLocation W43669997461 @default.
- W4366999746 hasOpenAccess W4366999746 @default.
- W4366999746 hasPrimaryLocation W43669997461 @default.
- W4366999746 hasRelatedWork W1480256999 @default.
- W4366999746 hasRelatedWork W1604847762 @default.
- W4366999746 hasRelatedWork W1824158299 @default.
- W4366999746 hasRelatedWork W1991120448 @default.
- W4366999746 hasRelatedWork W2001065678 @default.
- W4366999746 hasRelatedWork W2306328185 @default.
- W4366999746 hasRelatedWork W2314132665 @default.
- W4366999746 hasRelatedWork W2391384657 @default.
- W4366999746 hasRelatedWork W3170086649 @default.
- W4366999746 hasRelatedWork W97768505 @default.
- W4366999746 isParatext "false" @default.
- W4366999746 isRetracted "false" @default.
- W4366999746 workType "article" @default.