Matches in SemOpenAlex for { <https://semopenalex.org/work/W3163244872> ?p ?o ?g. }
Showing items 1 to 82 of
82
with 100 items per page.
- W3163244872 abstract "Neural network pruning with suitable retraining can yield networks with considerably fewer parameters than the original with comparable degrees of accuracy. Typical pruning methods require large, fully trained networks as a starting point from which they perform a time-intensive iterative pruning and retraining procedure to regain the original accuracy. We propose a novel pruning method that prunes a network real-time during training, reducing the overall training time to achieve an efficient compressed network. We introduce an activation density based analysis to identify the optimal relative sizing or compression for each layer of the network. Our method is architecture agnostic, allowing it to be employed on a wide variety of systems. For VGG-19 and ResNet18 on CIFAR-10, CIFAR-100, and TinyImageNet, we obtain exceedingly sparse networks (up to 200 x reduction in parameters and over 60 x reduction in inference compute operations in the best case) with accuracy comparable to the baseline network. By reducing the network size periodically during training, we achieve total training times that are shorter than those of previously proposed pruning methods. Furthermore, training compressed networks at different epochs with our proposed method yields considerable reduction in training compute complexity (1.6 x to 3.2 x lower) at near iso-accuracy as compared to a baseline network trained entirely from scratch." @default.
- W3163244872 created "2021-05-24" @default.
- W3163244872 creator A5050310538 @default.
- W3163244872 creator A5059671888 @default.
- W3163244872 creator A5079036687 @default.
- W3163244872 date "2021-01-10" @default.
- W3163244872 modified "2023-09-24" @default.
- W3163244872 title "Activation Density Driven Efficient Pruning in Training" @default.
- W3163244872 cites W2194775991 @default.
- W3163244872 cites W2585720638 @default.
- W3163244872 cites W2614143469 @default.
- W3163244872 cites W2798544842 @default.
- W3163244872 cites W2962851801 @default.
- W3163244872 cites W3091661482 @default.
- W3163244872 cites W4236868170 @default.
- W3163244872 cites W4240168186 @default.
- W3163244872 doi "https://doi.org/10.1109/icpr48806.2021.9413182" @default.
- W3163244872 hasPublicationYear "2021" @default.
- W3163244872 type Work @default.
- W3163244872 sameAs 3163244872 @default.
- W3163244872 citedByCount "1" @default.
- W3163244872 countsByYear W31632448722021 @default.
- W3163244872 crossrefType "proceedings-article" @default.
- W3163244872 hasAuthorship W3163244872A5050310538 @default.
- W3163244872 hasAuthorship W3163244872A5059671888 @default.
- W3163244872 hasAuthorship W3163244872A5079036687 @default.
- W3163244872 hasConcept C108010975 @default.
- W3163244872 hasConcept C111335779 @default.
- W3163244872 hasConcept C111368507 @default.
- W3163244872 hasConcept C119857082 @default.
- W3163244872 hasConcept C12725497 @default.
- W3163244872 hasConcept C127313418 @default.
- W3163244872 hasConcept C144133560 @default.
- W3163244872 hasConcept C154945302 @default.
- W3163244872 hasConcept C155202549 @default.
- W3163244872 hasConcept C193415008 @default.
- W3163244872 hasConcept C2524010 @default.
- W3163244872 hasConcept C2776214188 @default.
- W3163244872 hasConcept C2778712577 @default.
- W3163244872 hasConcept C33923547 @default.
- W3163244872 hasConcept C38652104 @default.
- W3163244872 hasConcept C41008148 @default.
- W3163244872 hasConcept C50644808 @default.
- W3163244872 hasConcept C6557445 @default.
- W3163244872 hasConcept C86803240 @default.
- W3163244872 hasConceptScore W3163244872C108010975 @default.
- W3163244872 hasConceptScore W3163244872C111335779 @default.
- W3163244872 hasConceptScore W3163244872C111368507 @default.
- W3163244872 hasConceptScore W3163244872C119857082 @default.
- W3163244872 hasConceptScore W3163244872C12725497 @default.
- W3163244872 hasConceptScore W3163244872C127313418 @default.
- W3163244872 hasConceptScore W3163244872C144133560 @default.
- W3163244872 hasConceptScore W3163244872C154945302 @default.
- W3163244872 hasConceptScore W3163244872C155202549 @default.
- W3163244872 hasConceptScore W3163244872C193415008 @default.
- W3163244872 hasConceptScore W3163244872C2524010 @default.
- W3163244872 hasConceptScore W3163244872C2776214188 @default.
- W3163244872 hasConceptScore W3163244872C2778712577 @default.
- W3163244872 hasConceptScore W3163244872C33923547 @default.
- W3163244872 hasConceptScore W3163244872C38652104 @default.
- W3163244872 hasConceptScore W3163244872C41008148 @default.
- W3163244872 hasConceptScore W3163244872C50644808 @default.
- W3163244872 hasConceptScore W3163244872C6557445 @default.
- W3163244872 hasConceptScore W3163244872C86803240 @default.
- W3163244872 hasFunder F4320306076 @default.
- W3163244872 hasLocation W31632448721 @default.
- W3163244872 hasOpenAccess W3163244872 @default.
- W3163244872 hasPrimaryLocation W31632448721 @default.
- W3163244872 hasRelatedWork W1513235864 @default.
- W3163244872 hasRelatedWork W2511279186 @default.
- W3163244872 hasRelatedWork W2963058055 @default.
- W3163244872 hasRelatedWork W3119508709 @default.
- W3163244872 hasRelatedWork W3189043577 @default.
- W3163244872 hasRelatedWork W3199608561 @default.
- W3163244872 hasRelatedWork W4206024927 @default.
- W3163244872 hasRelatedWork W4293102051 @default.
- W3163244872 hasRelatedWork W4312358390 @default.
- W3163244872 hasRelatedWork W1629725936 @default.
- W3163244872 isParatext "false" @default.
- W3163244872 isRetracted "false" @default.
- W3163244872 magId "3163244872" @default.
- W3163244872 workType "article" @default.