Matches in SemOpenAlex for { <https://semopenalex.org/work/W2928413811> ?p ?o ?g. }
Showing items 1 to 82 of
82
with 100 items per page.
- W2928413811 endingPage "16" @default.
- W2928413811 startingPage "1" @default.
- W2928413811 abstract "We provide here a novel method, called hypercolumn sparsification, to achieve high recognition performance for convolutional neural networks (CNNs) despite low-precision weights and activities during both training and test phases. This method is applicable to any CNN architecture that operates on signal patterns (e.g., audio, image, video) to extract information such as class membership. It operates on the stack of feature maps in each of the cascading feature matching and pooling layers through the processing hierarchy of the CNN by an explicit competitive process ( k -WTA, winner take all) that generates a sparse feature vector at each spatial location. This principle is inspired by local brain circuits, where neurons tuned to respond to different patterns in the incoming signals from an upstream region inhibit each other using interneurons, such that only the ones that are maximally activated survive the quenching threshold. We show this process of sparsification is critical for probabilistic learning of low-precision weights and bias terms, thereby making pattern recognition amenable for energy-efficient hardware implementations. Further, we show that hypercolumn sparsification could lead to more data-efficient learning as well as having an emergent property of significantly pruning down the number of connections in the network. A theoretical account and empirical analysis are provided to understand these effects better." @default.
- W2928413811 created "2019-04-11" @default.
- W2928413811 creator A5005563905 @default.
- W2928413811 creator A5055443197 @default.
- W2928413811 creator A5063930012 @default.
- W2928413811 creator A5069449307 @default.
- W2928413811 creator A5086401204 @default.
- W2928413811 date "2019-03-26" @default.
- W2928413811 modified "2023-10-03" @default.
- W2928413811 title "Hypercolumn Sparsification for Low-Power Convolutional Neural Networks" @default.
- W2928413811 cites W2004586880 @default.
- W2928413811 cites W2005876975 @default.
- W2928413811 cites W2054607054 @default.
- W2928413811 cites W2105464770 @default.
- W2928413811 cites W2126020484 @default.
- W2928413811 cites W2131097266 @default.
- W2928413811 cites W2132172482 @default.
- W2928413811 cites W2194775991 @default.
- W2928413811 cites W2300242332 @default.
- W2928413811 doi "https://doi.org/10.1145/3304104" @default.
- W2928413811 hasPublicationYear "2019" @default.
- W2928413811 type Work @default.
- W2928413811 sameAs 2928413811 @default.
- W2928413811 citedByCount "1" @default.
- W2928413811 countsByYear W29284138112021 @default.
- W2928413811 crossrefType "journal-article" @default.
- W2928413811 hasAuthorship W2928413811A5005563905 @default.
- W2928413811 hasAuthorship W2928413811A5055443197 @default.
- W2928413811 hasAuthorship W2928413811A5063930012 @default.
- W2928413811 hasAuthorship W2928413811A5069449307 @default.
- W2928413811 hasAuthorship W2928413811A5086401204 @default.
- W2928413811 hasBestOaLocation W29284138111 @default.
- W2928413811 hasConcept C108010975 @default.
- W2928413811 hasConcept C111919701 @default.
- W2928413811 hasConcept C11413529 @default.
- W2928413811 hasConcept C138885662 @default.
- W2928413811 hasConcept C153180895 @default.
- W2928413811 hasConcept C154945302 @default.
- W2928413811 hasConcept C2776401178 @default.
- W2928413811 hasConcept C41008148 @default.
- W2928413811 hasConcept C41895202 @default.
- W2928413811 hasConcept C6557445 @default.
- W2928413811 hasConcept C70437156 @default.
- W2928413811 hasConcept C81363708 @default.
- W2928413811 hasConcept C86803240 @default.
- W2928413811 hasConcept C98045186 @default.
- W2928413811 hasConceptScore W2928413811C108010975 @default.
- W2928413811 hasConceptScore W2928413811C111919701 @default.
- W2928413811 hasConceptScore W2928413811C11413529 @default.
- W2928413811 hasConceptScore W2928413811C138885662 @default.
- W2928413811 hasConceptScore W2928413811C153180895 @default.
- W2928413811 hasConceptScore W2928413811C154945302 @default.
- W2928413811 hasConceptScore W2928413811C2776401178 @default.
- W2928413811 hasConceptScore W2928413811C41008148 @default.
- W2928413811 hasConceptScore W2928413811C41895202 @default.
- W2928413811 hasConceptScore W2928413811C6557445 @default.
- W2928413811 hasConceptScore W2928413811C70437156 @default.
- W2928413811 hasConceptScore W2928413811C81363708 @default.
- W2928413811 hasConceptScore W2928413811C86803240 @default.
- W2928413811 hasConceptScore W2928413811C98045186 @default.
- W2928413811 hasFunder F4320332180 @default.
- W2928413811 hasIssue "2" @default.
- W2928413811 hasLocation W29284138111 @default.
- W2928413811 hasOpenAccess W2928413811 @default.
- W2928413811 hasPrimaryLocation W29284138111 @default.
- W2928413811 hasRelatedWork W2319888919 @default.
- W2928413811 hasRelatedWork W2514274290 @default.
- W2928413811 hasRelatedWork W2517027266 @default.
- W2928413811 hasRelatedWork W2613736958 @default.
- W2928413811 hasRelatedWork W2758063741 @default.
- W2928413811 hasRelatedWork W2760085659 @default.
- W2928413811 hasRelatedWork W2765989371 @default.
- W2928413811 hasRelatedWork W2792080776 @default.
- W2928413811 hasRelatedWork W2969680539 @default.
- W2928413811 hasRelatedWork W4283454150 @default.
- W2928413811 hasVolume "15" @default.
- W2928413811 isParatext "false" @default.
- W2928413811 isRetracted "false" @default.
- W2928413811 magId "2928413811" @default.
- W2928413811 workType "article" @default.