Matches in SemOpenAlex for { <https://semopenalex.org/work/W3128194313> ?p ?o ?g. }
Showing items 1 to 100 of
100
with 100 items per page.
- W3128194313 abstract "Sparsifying deep neural networks is of paramount interest in many areas, especially when those networks have to be implemented on low-memory devices. In this article, we propose a new formulation of the problem of generating sparse weights for a neural network. By leveraging the properties of standard nonlinear activation functions, we show that the problem is equivalent to an approximate subdifferential inclusion problem. The accuracy of the approximation controls the sparsity. We show that the proposed approach is valid for a broad class of activation functions (ReLU, sigmoid, softmax). We propose an iterative optimization algorithm to induce sparsity whose convergence is guaranteed. Because of the algorithm flexibility, the sparsity can be ensured from partial training data in a minibatch manner. To demonstrate the effectiveness of our method, we perform experiments on various networks in different applicative contexts: image classification, speech recognition, natural language processing, and time-series forecasting." @default.
- W3128194313 created "2021-02-15" @default.
- W3128194313 creator A5012458684 @default.
- W3128194313 creator A5062845079 @default.
- W3128194313 date "2021-07-18" @default.
- W3128194313 modified "2023-10-15" @default.
- W3128194313 title "Sparsifying Networks via Subdifferential Inclusion" @default.
- W3128194313 cites W1494198834 @default.
- W3128194313 cites W1686810756 @default.
- W3128194313 cites W2019569173 @default.
- W3128194313 cites W205960364 @default.
- W3128194313 cites W2096764219 @default.
- W3128194313 cites W2114766824 @default.
- W3128194313 cites W2115321886 @default.
- W3128194313 cites W2134273960 @default.
- W3128194313 cites W2142280715 @default.
- W3128194313 cites W2156150815 @default.
- W3128194313 cites W2194775991 @default.
- W3128194313 cites W2525332836 @default.
- W3128194313 cites W2619890685 @default.
- W3128194313 cites W2768501777 @default.
- W3128194313 cites W2892009249 @default.
- W3128194313 cites W2907886210 @default.
- W3128194313 cites W2951569836 @default.
- W3128194313 cites W2952746978 @default.
- W3128194313 cites W2962780374 @default.
- W3128194313 cites W2963000224 @default.
- W3128194313 cites W2963117513 @default.
- W3128194313 cites W2963247446 @default.
- W3128194313 cites W2963446712 @default.
- W3128194313 cites W2963674932 @default.
- W3128194313 cites W2963766446 @default.
- W3128194313 cites W2963813662 @default.
- W3128194313 cites W2963828549 @default.
- W3128194313 cites W2963918968 @default.
- W3128194313 cites W2963991999 @default.
- W3128194313 cites W2964110616 @default.
- W3128194313 cites W2964217527 @default.
- W3128194313 cites W2964233199 @default.
- W3128194313 cites W2964259004 @default.
- W3128194313 cites W2970860468 @default.
- W3128194313 cites W2973215447 @default.
- W3128194313 cites W2982083293 @default.
- W3128194313 cites W2995463996 @default.
- W3128194313 cites W2995492258 @default.
- W3128194313 cites W2995816250 @default.
- W3128194313 cites W2996577930 @default.
- W3128194313 cites W3005273253 @default.
- W3128194313 cites W3006592723 @default.
- W3128194313 cites W3009043942 @default.
- W3128194313 cites W3035180000 @default.
- W3128194313 cites W3038041907 @default.
- W3128194313 cites W3044604993 @default.
- W3128194313 cites W3115511450 @default.
- W3128194313 hasPublicationYear "2021" @default.
- W3128194313 type Work @default.
- W3128194313 sameAs 3128194313 @default.
- W3128194313 citedByCount "1" @default.
- W3128194313 countsByYear W31281943132021 @default.
- W3128194313 crossrefType "proceedings-article" @default.
- W3128194313 hasAuthorship W3128194313A5012458684 @default.
- W3128194313 hasAuthorship W3128194313A5062845079 @default.
- W3128194313 hasBestOaLocation W31281943131 @default.
- W3128194313 hasConcept C107993555 @default.
- W3128194313 hasConcept C109359841 @default.
- W3128194313 hasConcept C112680207 @default.
- W3128194313 hasConcept C126255220 @default.
- W3128194313 hasConcept C144024400 @default.
- W3128194313 hasConcept C157972887 @default.
- W3128194313 hasConcept C200661725 @default.
- W3128194313 hasConcept C2524010 @default.
- W3128194313 hasConcept C33923547 @default.
- W3128194313 hasConcept C41008148 @default.
- W3128194313 hasConceptScore W3128194313C107993555 @default.
- W3128194313 hasConceptScore W3128194313C109359841 @default.
- W3128194313 hasConceptScore W3128194313C112680207 @default.
- W3128194313 hasConceptScore W3128194313C126255220 @default.
- W3128194313 hasConceptScore W3128194313C144024400 @default.
- W3128194313 hasConceptScore W3128194313C157972887 @default.
- W3128194313 hasConceptScore W3128194313C200661725 @default.
- W3128194313 hasConceptScore W3128194313C2524010 @default.
- W3128194313 hasConceptScore W3128194313C33923547 @default.
- W3128194313 hasConceptScore W3128194313C41008148 @default.
- W3128194313 hasLocation W31281943131 @default.
- W3128194313 hasOpenAccess W3128194313 @default.
- W3128194313 hasPrimaryLocation W31281943131 @default.
- W3128194313 hasRelatedWork W1985683962 @default.
- W3128194313 hasRelatedWork W1990342248 @default.
- W3128194313 hasRelatedWork W2048623688 @default.
- W3128194313 hasRelatedWork W2057003034 @default.
- W3128194313 hasRelatedWork W2060409023 @default.
- W3128194313 hasRelatedWork W2795367102 @default.
- W3128194313 hasRelatedWork W2882986165 @default.
- W3128194313 hasRelatedWork W3134741060 @default.
- W3128194313 hasRelatedWork W3186178614 @default.
- W3128194313 hasRelatedWork W4250318239 @default.
- W3128194313 isParatext "false" @default.
- W3128194313 isRetracted "false" @default.
- W3128194313 magId "3128194313" @default.
- W3128194313 workType "article" @default.