Matches in SemOpenAlex for { <https://semopenalex.org/work/W2407624399> ?p ?o ?g. }
Showing items 1 to 88 of
88
with 100 items per page.
- W2407624399 endingPage "649" @default.
- W2407624399 startingPage "638" @default.
- W2407624399 abstract "Convolutional neural networks (CNNs) are one of the most successful deep architectures in machine learning. While they achieve superior recognition rate, the intensive computation of CNNs limits their applicability. In this paper, we propose a method based on separable filters to reduce the computational cost. By using Singular Value Decompositions (SVDs), a 2D filter in the CNNs can be approximated by the product of two 1D filters, and the 2D convolution can be computed via two consecutive 1D convolutions. We implemented a batched SVD routine on GPUs that can compute the SVD of multiple small matrices simultaneously, and three convolution methods using different memory spaces according to the filter size. Comparing to the state-of-art GPU implementations of CNNs, experimental results show that our methods can achieve up to 2.66 times speedup in the forward pass and up to 2.35 times speedup in the backward pass." @default.
- W2407624399 created "2016-06-24" @default.
- W2407624399 creator A5040718356 @default.
- W2407624399 creator A5058589126 @default.
- W2407624399 date "2015-01-01" @default.
- W2407624399 modified "2023-09-24" @default.
- W2407624399 title "Improving Performance of Convolutional Neural Networks by Separable Filters on GPU" @default.
- W2407624399 cites W1996901117 @default.
- W2407624399 cites W2000723898 @default.
- W2407624399 cites W2004026774 @default.
- W2407624399 cites W2047614366 @default.
- W2407624399 cites W2112796928 @default.
- W2407624399 cites W2120480077 @default.
- W2407624399 cites W2121775913 @default.
- W2407624399 cites W2145287260 @default.
- W2407624399 cites W4229666556 @default.
- W2407624399 cites W4252780833 @default.
- W2407624399 cites W77200240 @default.
- W2407624399 doi "https://doi.org/10.1007/978-3-662-48096-0_49" @default.
- W2407624399 hasPublicationYear "2015" @default.
- W2407624399 type Work @default.
- W2407624399 sameAs 2407624399 @default.
- W2407624399 citedByCount "4" @default.
- W2407624399 countsByYear W24076243992017 @default.
- W2407624399 countsByYear W24076243992018 @default.
- W2407624399 countsByYear W24076243992020 @default.
- W2407624399 countsByYear W24076243992022 @default.
- W2407624399 crossrefType "book-chapter" @default.
- W2407624399 hasAuthorship W2407624399A5040718356 @default.
- W2407624399 hasAuthorship W2407624399A5058589126 @default.
- W2407624399 hasBestOaLocation W24076243991 @default.
- W2407624399 hasConcept C106131492 @default.
- W2407624399 hasConcept C11413529 @default.
- W2407624399 hasConcept C114614502 @default.
- W2407624399 hasConcept C134306372 @default.
- W2407624399 hasConcept C153180895 @default.
- W2407624399 hasConcept C154945302 @default.
- W2407624399 hasConcept C173608175 @default.
- W2407624399 hasConcept C22789450 @default.
- W2407624399 hasConcept C31972630 @default.
- W2407624399 hasConcept C33923547 @default.
- W2407624399 hasConcept C41008148 @default.
- W2407624399 hasConcept C45347329 @default.
- W2407624399 hasConcept C45374587 @default.
- W2407624399 hasConcept C459310 @default.
- W2407624399 hasConcept C50644808 @default.
- W2407624399 hasConcept C68339613 @default.
- W2407624399 hasConcept C70710897 @default.
- W2407624399 hasConcept C74193536 @default.
- W2407624399 hasConcept C81363708 @default.
- W2407624399 hasConceptScore W2407624399C106131492 @default.
- W2407624399 hasConceptScore W2407624399C11413529 @default.
- W2407624399 hasConceptScore W2407624399C114614502 @default.
- W2407624399 hasConceptScore W2407624399C134306372 @default.
- W2407624399 hasConceptScore W2407624399C153180895 @default.
- W2407624399 hasConceptScore W2407624399C154945302 @default.
- W2407624399 hasConceptScore W2407624399C173608175 @default.
- W2407624399 hasConceptScore W2407624399C22789450 @default.
- W2407624399 hasConceptScore W2407624399C31972630 @default.
- W2407624399 hasConceptScore W2407624399C33923547 @default.
- W2407624399 hasConceptScore W2407624399C41008148 @default.
- W2407624399 hasConceptScore W2407624399C45347329 @default.
- W2407624399 hasConceptScore W2407624399C45374587 @default.
- W2407624399 hasConceptScore W2407624399C459310 @default.
- W2407624399 hasConceptScore W2407624399C50644808 @default.
- W2407624399 hasConceptScore W2407624399C68339613 @default.
- W2407624399 hasConceptScore W2407624399C70710897 @default.
- W2407624399 hasConceptScore W2407624399C74193536 @default.
- W2407624399 hasConceptScore W2407624399C81363708 @default.
- W2407624399 hasLocation W24076243991 @default.
- W2407624399 hasOpenAccess W2407624399 @default.
- W2407624399 hasPrimaryLocation W24076243991 @default.
- W2407624399 hasRelatedWork W156843270 @default.
- W2407624399 hasRelatedWork W2042026112 @default.
- W2407624399 hasRelatedWork W2757410581 @default.
- W2407624399 hasRelatedWork W2767899175 @default.
- W2407624399 hasRelatedWork W2770717529 @default.
- W2407624399 hasRelatedWork W2796942851 @default.
- W2407624399 hasRelatedWork W2949649078 @default.
- W2407624399 hasRelatedWork W2963993660 @default.
- W2407624399 hasRelatedWork W2989991345 @default.
- W2407624399 hasRelatedWork W3004235922 @default.
- W2407624399 isParatext "false" @default.
- W2407624399 isRetracted "false" @default.
- W2407624399 magId "2407624399" @default.
- W2407624399 workType "book-chapter" @default.