Matches in SemOpenAlex for { <https://semopenalex.org/work/W2605195840> ?p ?o ?g. }
Showing items 1 to 68 of
68
with 100 items per page.
- W2605195840 abstract "The driving force behind deep networks is their ability to compactly represent rich classes of functions. The primary notion for formally reasoning about this phenomenon is expressive efficiency, which refers to a situation where one network must grow unfeasibly large in order to realize (or approximate) functions of another. To date, expressive efficiency analyses focused on the architectural feature of depth, showing that deep networks are representationally superior to shallow ones. In this paper we study the expressive efficiency brought forth by connectivity, motivated by the observation that modern networks interconnect their layers in elaborate ways. We focus on dilated convolutional networks, a family of deep models delivering state of the art performance in sequence processing tasks. By introducing and analyzing the concept of mixed tensor decompositions, we prove that interconnecting dilated convolutional networks can lead to expressive efficiency. In particular, we show that even a single connection between intermediate layers can already lead to an almost quadratic gap, which in large-scale settings typically makes the difference between a model that is practical and one that is not. Empirical evaluation demonstrates how the expressive efficiency of connectivity, similarly to that of depth, translates into gains in accuracy. This leads us to believe that expressive efficiency may serve a key role in the development of new tools for deep network design." @default.
- W2605195840 created "2017-04-14" @default.
- W2605195840 creator A5029113299 @default.
- W2605195840 creator A5060414926 @default.
- W2605195840 creator A5075004917 @default.
- W2605195840 date "2017-03-01" @default.
- W2605195840 modified "2023-09-28" @default.
- W2605195840 title "Boosting Dilated Convolutional Networks with Mixed Tensor Decompositions" @default.
- W2605195840 hasPublicationYear "2017" @default.
- W2605195840 type Work @default.
- W2605195840 sameAs 2605195840 @default.
- W2605195840 citedByCount "3" @default.
- W2605195840 countsByYear W26051958402019 @default.
- W2605195840 countsByYear W26051958402020 @default.
- W2605195840 countsByYear W26051958402021 @default.
- W2605195840 crossrefType "proceedings-article" @default.
- W2605195840 hasAuthorship W2605195840A5029113299 @default.
- W2605195840 hasAuthorship W2605195840A5060414926 @default.
- W2605195840 hasAuthorship W2605195840A5075004917 @default.
- W2605195840 hasConcept C108583219 @default.
- W2605195840 hasConcept C123745756 @default.
- W2605195840 hasConcept C154945302 @default.
- W2605195840 hasConcept C155281189 @default.
- W2605195840 hasConcept C174348530 @default.
- W2605195840 hasConcept C202444582 @default.
- W2605195840 hasConcept C31258907 @default.
- W2605195840 hasConcept C33923547 @default.
- W2605195840 hasConcept C41008148 @default.
- W2605195840 hasConcept C46686674 @default.
- W2605195840 hasConcept C80444323 @default.
- W2605195840 hasConceptScore W2605195840C108583219 @default.
- W2605195840 hasConceptScore W2605195840C123745756 @default.
- W2605195840 hasConceptScore W2605195840C154945302 @default.
- W2605195840 hasConceptScore W2605195840C155281189 @default.
- W2605195840 hasConceptScore W2605195840C174348530 @default.
- W2605195840 hasConceptScore W2605195840C202444582 @default.
- W2605195840 hasConceptScore W2605195840C31258907 @default.
- W2605195840 hasConceptScore W2605195840C33923547 @default.
- W2605195840 hasConceptScore W2605195840C41008148 @default.
- W2605195840 hasConceptScore W2605195840C46686674 @default.
- W2605195840 hasConceptScore W2605195840C80444323 @default.
- W2605195840 hasLocation W26051958401 @default.
- W2605195840 hasOpenAccess W2605195840 @default.
- W2605195840 hasPrimaryLocation W26051958401 @default.
- W2605195840 hasRelatedWork W1493372406 @default.
- W2605195840 hasRelatedWork W1600323995 @default.
- W2605195840 hasRelatedWork W2164700406 @default.
- W2605195840 hasRelatedWork W2224525916 @default.
- W2605195840 hasRelatedWork W2276387674 @default.
- W2605195840 hasRelatedWork W2534135591 @default.
- W2605195840 hasRelatedWork W2766953637 @default.
- W2605195840 hasRelatedWork W2786262858 @default.
- W2605195840 hasRelatedWork W2803911450 @default.
- W2605195840 hasRelatedWork W2893795152 @default.
- W2605195840 hasRelatedWork W2963833878 @default.
- W2605195840 hasRelatedWork W2978279988 @default.
- W2605195840 hasRelatedWork W2982128187 @default.
- W2605195840 hasRelatedWork W2995720879 @default.
- W2605195840 hasRelatedWork W3009352062 @default.
- W2605195840 hasRelatedWork W3035980362 @default.
- W2605195840 hasRelatedWork W3087641101 @default.
- W2605195840 hasRelatedWork W3125151176 @default.
- W2605195840 hasRelatedWork W3182810299 @default.
- W2605195840 hasRelatedWork W3187108119 @default.
- W2605195840 isParatext "false" @default.
- W2605195840 isRetracted "false" @default.
- W2605195840 magId "2605195840" @default.
- W2605195840 workType "article" @default.