Matches in SemOpenAlex for { <https://semopenalex.org/work/W2971337223> ?p ?o ?g. }
Showing items 1 to 87 of
87
with 100 items per page.
- W2971337223 abstract "Pooling is used in most of deep convolutional neural networks as a feature downsampling method, in order to reduce computation complexity and increase the receptive field size. Since the traditional max/average pooling layers tend to cause severe information loss, spectral pooling with discrete Fourier transform (DFT) is considered as an advisable alternative. In this paper, we propose a novel 2D-discrete cosine transform (2D-DCT) based pooling method for deep neural networks. Due to the energy compaction property, DCT pooling preserves considerably more information than DFT. Moreover, inspired by the separability of DCT, we precompute the transform matrices and embed them into linear layers for parallelization and acceleration on GPUs. Experimental results indicate that the proposed DCT based pooling layer outperforms previous pooling methods on multiple image classification datasets with negligible extra time consumption." @default.
- W2971337223 created "2019-09-05" @default.
- W2971337223 creator A5029688192 @default.
- W2971337223 creator A5050229964 @default.
- W2971337223 date "2019-09-01" @default.
- W2971337223 modified "2023-09-30" @default.
- W2971337223 title "DCT Based Information-Preserving Pooling for Deep Neural Networks" @default.
- W2971337223 cites W1840106123 @default.
- W2971337223 cites W2031614119 @default.
- W2971337223 cites W2061171222 @default.
- W2971337223 cites W2097117768 @default.
- W2971337223 cites W2194775991 @default.
- W2971337223 cites W2559156603 @default.
- W2971337223 cites W2802252728 @default.
- W2971337223 cites W2896786907 @default.
- W2971337223 cites W2964012402 @default.
- W2971337223 doi "https://doi.org/10.1109/icip.2019.8802962" @default.
- W2971337223 hasPublicationYear "2019" @default.
- W2971337223 type Work @default.
- W2971337223 sameAs 2971337223 @default.
- W2971337223 citedByCount "4" @default.
- W2971337223 countsByYear W29713372232020 @default.
- W2971337223 countsByYear W29713372232021 @default.
- W2971337223 countsByYear W29713372232022 @default.
- W2971337223 countsByYear W29713372232023 @default.
- W2971337223 crossrefType "proceedings-article" @default.
- W2971337223 hasAuthorship W2971337223A5029688192 @default.
- W2971337223 hasAuthorship W2971337223A5050229964 @default.
- W2971337223 hasConcept C102519508 @default.
- W2971337223 hasConcept C110384440 @default.
- W2971337223 hasConcept C11413529 @default.
- W2971337223 hasConcept C115961682 @default.
- W2971337223 hasConcept C134306372 @default.
- W2971337223 hasConcept C153180895 @default.
- W2971337223 hasConcept C154945302 @default.
- W2971337223 hasConcept C203024314 @default.
- W2971337223 hasConcept C2221639 @default.
- W2971337223 hasConcept C33923547 @default.
- W2971337223 hasConcept C41008148 @default.
- W2971337223 hasConcept C50644808 @default.
- W2971337223 hasConcept C57733114 @default.
- W2971337223 hasConcept C70437156 @default.
- W2971337223 hasConcept C76563020 @default.
- W2971337223 hasConcept C81363708 @default.
- W2971337223 hasConceptScore W2971337223C102519508 @default.
- W2971337223 hasConceptScore W2971337223C110384440 @default.
- W2971337223 hasConceptScore W2971337223C11413529 @default.
- W2971337223 hasConceptScore W2971337223C115961682 @default.
- W2971337223 hasConceptScore W2971337223C134306372 @default.
- W2971337223 hasConceptScore W2971337223C153180895 @default.
- W2971337223 hasConceptScore W2971337223C154945302 @default.
- W2971337223 hasConceptScore W2971337223C203024314 @default.
- W2971337223 hasConceptScore W2971337223C2221639 @default.
- W2971337223 hasConceptScore W2971337223C33923547 @default.
- W2971337223 hasConceptScore W2971337223C41008148 @default.
- W2971337223 hasConceptScore W2971337223C50644808 @default.
- W2971337223 hasConceptScore W2971337223C57733114 @default.
- W2971337223 hasConceptScore W2971337223C70437156 @default.
- W2971337223 hasConceptScore W2971337223C76563020 @default.
- W2971337223 hasConceptScore W2971337223C81363708 @default.
- W2971337223 hasLocation W29713372231 @default.
- W2971337223 hasOpenAccess W2971337223 @default.
- W2971337223 hasPrimaryLocation W29713372231 @default.
- W2971337223 hasRelatedWork W1499713894 @default.
- W2971337223 hasRelatedWork W1509715878 @default.
- W2971337223 hasRelatedWork W1529465528 @default.
- W2971337223 hasRelatedWork W1971320633 @default.
- W2971337223 hasRelatedWork W1980850744 @default.
- W2971337223 hasRelatedWork W2039205241 @default.
- W2971337223 hasRelatedWork W2043463146 @default.
- W2971337223 hasRelatedWork W2127793713 @default.
- W2971337223 hasRelatedWork W2162639497 @default.
- W2971337223 hasRelatedWork W2167970954 @default.
- W2971337223 hasRelatedWork W2364851361 @default.
- W2971337223 hasRelatedWork W2609675817 @default.
- W2971337223 hasRelatedWork W2783569857 @default.
- W2971337223 hasRelatedWork W2787457768 @default.
- W2971337223 hasRelatedWork W2889398290 @default.
- W2971337223 hasRelatedWork W2889827362 @default.
- W2971337223 hasRelatedWork W2949069042 @default.
- W2971337223 hasRelatedWork W3142639079 @default.
- W2971337223 hasRelatedWork W1999843301 @default.
- W2971337223 hasRelatedWork W2114928149 @default.
- W2971337223 isParatext "false" @default.
- W2971337223 isRetracted "false" @default.
- W2971337223 magId "2971337223" @default.
- W2971337223 workType "article" @default.