Matches in SemOpenAlex for { <https://semopenalex.org/work/W3100643896> ?p ?o ?g. }
Showing items 1 to 88 of
88
with 100 items per page.
- W3100643896 endingPage "19237" @default.
- W3100643896 startingPage "19224" @default.
- W3100643896 abstract "We introduce a new theoretical framework to analyze deep learning optimization with connection to its generalization error. Existing frameworks such as mean field theory and neural tangent kernel theory for neural network optimization analysis typically require taking limit of infinite width of the network to show its global convergence. This potentially makes it difficult to directly deal with finite width network; especially in the neural tangent kernel regime, we cannot reveal favorable properties of neural networks beyond kernel methods. To realize more natural analysis, we consider a completely different approach in which we formulate the parameter training as a transportation map estimation and show its global convergence via the theory of the infinite dimensional Langevin dynamics. This enables us to analyze narrow and wide networks in a unifying manner. Moreover, we give generalization gap and excess risk bounds for the solution obtained by the dynamics. The excess risk bound achieves the so-called fast learning rate. In particular, we show an exponential convergence for a classification problem and a minimax optimal rate for a regression problem." @default.
- W3100643896 created "2020-11-23" @default.
- W3100643896 creator A5078812767 @default.
- W3100643896 date "2020-01-01" @default.
- W3100643896 modified "2023-09-29" @default.
- W3100643896 title "Generalization bound of globally optimal non-convex neural network training: Transportation map estimation by infinite dimensional Langevin dynamics" @default.
- W3100643896 hasPublicationYear "2020" @default.
- W3100643896 type Work @default.
- W3100643896 sameAs 3100643896 @default.
- W3100643896 citedByCount "8" @default.
- W3100643896 countsByYear W31006438962020 @default.
- W3100643896 countsByYear W31006438962021 @default.
- W3100643896 crossrefType "proceedings-article" @default.
- W3100643896 hasAuthorship W3100643896A5078812767 @default.
- W3100643896 hasConcept C105795698 @default.
- W3100643896 hasConcept C118615104 @default.
- W3100643896 hasConcept C126255220 @default.
- W3100643896 hasConcept C127162648 @default.
- W3100643896 hasConcept C134306372 @default.
- W3100643896 hasConcept C138187205 @default.
- W3100643896 hasConcept C149728462 @default.
- W3100643896 hasConcept C151201525 @default.
- W3100643896 hasConcept C154945302 @default.
- W3100643896 hasConcept C162324750 @default.
- W3100643896 hasConcept C177148314 @default.
- W3100643896 hasConcept C2524010 @default.
- W3100643896 hasConcept C2777303404 @default.
- W3100643896 hasConcept C2780004032 @default.
- W3100643896 hasConcept C28826006 @default.
- W3100643896 hasConcept C31258907 @default.
- W3100643896 hasConcept C33923547 @default.
- W3100643896 hasConcept C41008148 @default.
- W3100643896 hasConcept C50522688 @default.
- W3100643896 hasConcept C50644808 @default.
- W3100643896 hasConcept C57869625 @default.
- W3100643896 hasConcept C74193536 @default.
- W3100643896 hasConceptScore W3100643896C105795698 @default.
- W3100643896 hasConceptScore W3100643896C118615104 @default.
- W3100643896 hasConceptScore W3100643896C126255220 @default.
- W3100643896 hasConceptScore W3100643896C127162648 @default.
- W3100643896 hasConceptScore W3100643896C134306372 @default.
- W3100643896 hasConceptScore W3100643896C138187205 @default.
- W3100643896 hasConceptScore W3100643896C149728462 @default.
- W3100643896 hasConceptScore W3100643896C151201525 @default.
- W3100643896 hasConceptScore W3100643896C154945302 @default.
- W3100643896 hasConceptScore W3100643896C162324750 @default.
- W3100643896 hasConceptScore W3100643896C177148314 @default.
- W3100643896 hasConceptScore W3100643896C2524010 @default.
- W3100643896 hasConceptScore W3100643896C2777303404 @default.
- W3100643896 hasConceptScore W3100643896C2780004032 @default.
- W3100643896 hasConceptScore W3100643896C28826006 @default.
- W3100643896 hasConceptScore W3100643896C31258907 @default.
- W3100643896 hasConceptScore W3100643896C33923547 @default.
- W3100643896 hasConceptScore W3100643896C41008148 @default.
- W3100643896 hasConceptScore W3100643896C50522688 @default.
- W3100643896 hasConceptScore W3100643896C50644808 @default.
- W3100643896 hasConceptScore W3100643896C57869625 @default.
- W3100643896 hasConceptScore W3100643896C74193536 @default.
- W3100643896 hasLocation W31006438961 @default.
- W3100643896 hasOpenAccess W3100643896 @default.
- W3100643896 hasPrimaryLocation W31006438961 @default.
- W3100643896 hasRelatedWork W2595751897 @default.
- W3100643896 hasRelatedWork W2626283943 @default.
- W3100643896 hasRelatedWork W2752253833 @default.
- W3100643896 hasRelatedWork W2798989470 @default.
- W3100643896 hasRelatedWork W2809090039 @default.
- W3100643896 hasRelatedWork W2896520257 @default.
- W3100643896 hasRelatedWork W2921395510 @default.
- W3100643896 hasRelatedWork W2943540325 @default.
- W3100643896 hasRelatedWork W2970723196 @default.
- W3100643896 hasRelatedWork W2973114996 @default.
- W3100643896 hasRelatedWork W2999073519 @default.
- W3100643896 hasRelatedWork W3012339907 @default.
- W3100643896 hasRelatedWork W3041002672 @default.
- W3100643896 hasRelatedWork W3041957046 @default.
- W3100643896 hasRelatedWork W3082045641 @default.
- W3100643896 hasRelatedWork W3091853547 @default.
- W3100643896 hasRelatedWork W3097149416 @default.
- W3100643896 hasRelatedWork W3121423254 @default.
- W3100643896 hasRelatedWork W3134671887 @default.
- W3100643896 hasRelatedWork W3202491425 @default.
- W3100643896 hasVolume "33" @default.
- W3100643896 isParatext "false" @default.
- W3100643896 isRetracted "false" @default.
- W3100643896 magId "3100643896" @default.
- W3100643896 workType "article" @default.