Matches in SemOpenAlex for { <https://semopenalex.org/work/W3213653485> ?p ?o ?g. }
Showing items 1 to 87 of
87
with 100 items per page.
- W3213653485 abstract "Implicit neural networks, a.k.a., deep equilibrium networks, are a class of implicit-depth learning models where function evaluation is performed by solving a fixed point equation. They generalize classic feedforward models and are equivalent to infinite-depth weight-tied feedforward networks. While implicit models show improved accuracy and significant reduction in memory consumption, they can suffer from ill-posedness and convergence instability. This paper provides a new framework to design well-posed and robust implicit neural networks based upon contraction theory for the non-Euclidean norm $ell_infty$. Our framework includes (i) a novel condition for well-posedness based on one-sided Lipschitz constants, (ii) an average iteration for computing fixed-points, and (iii) explicit estimates on input-output Lipschitz constants. Additionally, we design a training problem with the well-posedness condition and the average iteration as constraints and, to achieve robust models, with the input-output Lipschitz constant as a regularizer. Our $ell_infty$ well-posedness condition leads to a larger polytopic training search space than existing conditions and our average iteration enjoys accelerated convergence. Finally, we perform several numerical experiments for function estimation and digit classification through the MNIST data set. Our numerical results demonstrate improved accuracy and robustness of the implicit models with smaller input-output Lipschitz bounds." @default.
- W3213653485 created "2021-11-22" @default.
- W3213653485 creator A5008697351 @default.
- W3213653485 creator A5014088551 @default.
- W3213653485 creator A5059595265 @default.
- W3213653485 creator A5068353454 @default.
- W3213653485 date "2021-12-06" @default.
- W3213653485 modified "2023-10-18" @default.
- W3213653485 title "Robust Implicit Networks via Non-Euclidean Contractions" @default.
- W3213653485 hasPublicationYear "2021" @default.
- W3213653485 type Work @default.
- W3213653485 sameAs 3213653485 @default.
- W3213653485 citedByCount "1" @default.
- W3213653485 countsByYear W32136534852021 @default.
- W3213653485 crossrefType "proceedings-article" @default.
- W3213653485 hasAuthorship W3213653485A5008697351 @default.
- W3213653485 hasAuthorship W3213653485A5014088551 @default.
- W3213653485 hasAuthorship W3213653485A5059595265 @default.
- W3213653485 hasAuthorship W3213653485A5068353454 @default.
- W3213653485 hasConcept C104317684 @default.
- W3213653485 hasConcept C126255220 @default.
- W3213653485 hasConcept C134306372 @default.
- W3213653485 hasConcept C154945302 @default.
- W3213653485 hasConcept C162324750 @default.
- W3213653485 hasConcept C17744445 @default.
- W3213653485 hasConcept C185592680 @default.
- W3213653485 hasConcept C190502265 @default.
- W3213653485 hasConcept C191795146 @default.
- W3213653485 hasConcept C199539241 @default.
- W3213653485 hasConcept C22324862 @default.
- W3213653485 hasConcept C2777303404 @default.
- W3213653485 hasConcept C28826006 @default.
- W3213653485 hasConcept C33923547 @default.
- W3213653485 hasConcept C41008148 @default.
- W3213653485 hasConcept C50522688 @default.
- W3213653485 hasConcept C50644808 @default.
- W3213653485 hasConcept C55493867 @default.
- W3213653485 hasConcept C61445026 @default.
- W3213653485 hasConcept C63479239 @default.
- W3213653485 hasConceptScore W3213653485C104317684 @default.
- W3213653485 hasConceptScore W3213653485C126255220 @default.
- W3213653485 hasConceptScore W3213653485C134306372 @default.
- W3213653485 hasConceptScore W3213653485C154945302 @default.
- W3213653485 hasConceptScore W3213653485C162324750 @default.
- W3213653485 hasConceptScore W3213653485C17744445 @default.
- W3213653485 hasConceptScore W3213653485C185592680 @default.
- W3213653485 hasConceptScore W3213653485C190502265 @default.
- W3213653485 hasConceptScore W3213653485C191795146 @default.
- W3213653485 hasConceptScore W3213653485C199539241 @default.
- W3213653485 hasConceptScore W3213653485C22324862 @default.
- W3213653485 hasConceptScore W3213653485C2777303404 @default.
- W3213653485 hasConceptScore W3213653485C28826006 @default.
- W3213653485 hasConceptScore W3213653485C33923547 @default.
- W3213653485 hasConceptScore W3213653485C41008148 @default.
- W3213653485 hasConceptScore W3213653485C50522688 @default.
- W3213653485 hasConceptScore W3213653485C50644808 @default.
- W3213653485 hasConceptScore W3213653485C55493867 @default.
- W3213653485 hasConceptScore W3213653485C61445026 @default.
- W3213653485 hasConceptScore W3213653485C63479239 @default.
- W3213653485 hasLocation W32136534851 @default.
- W3213653485 hasOpenAccess W3213653485 @default.
- W3213653485 hasPrimaryLocation W32136534851 @default.
- W3213653485 hasRelatedWork W1489077689 @default.
- W3213653485 hasRelatedWork W1971728746 @default.
- W3213653485 hasRelatedWork W2027286734 @default.
- W3213653485 hasRelatedWork W2055828890 @default.
- W3213653485 hasRelatedWork W2115849850 @default.
- W3213653485 hasRelatedWork W2156779765 @default.
- W3213653485 hasRelatedWork W2734748247 @default.
- W3213653485 hasRelatedWork W2771061327 @default.
- W3213653485 hasRelatedWork W2950967979 @default.
- W3213653485 hasRelatedWork W2953571613 @default.
- W3213653485 hasRelatedWork W2979411530 @default.
- W3213653485 hasRelatedWork W3012021965 @default.
- W3213653485 hasRelatedWork W3089190316 @default.
- W3213653485 hasRelatedWork W3106830411 @default.
- W3213653485 hasRelatedWork W3134012451 @default.
- W3213653485 hasRelatedWork W3145662549 @default.
- W3213653485 hasRelatedWork W3166578625 @default.
- W3213653485 hasRelatedWork W3186206582 @default.
- W3213653485 hasRelatedWork W3200449492 @default.
- W3213653485 hasRelatedWork W42963109 @default.
- W3213653485 hasVolume "34" @default.
- W3213653485 isParatext "false" @default.
- W3213653485 isRetracted "false" @default.
- W3213653485 magId "3213653485" @default.
- W3213653485 workType "article" @default.