Matches in SemOpenAlex for { <https://semopenalex.org/work/W3100848797> ?p ?o ?g. }
- W3100848797 abstract "Filter pruning is widely used to reduce the computation of deep learning, enabling the deployment of Deep Neural Networks (DNNs) in resource-limited devices. Conventional Hard Filter Pruning (HFP) method zeroizes pruned filters and stops updating them, thus reducing the search space of the model. On the contrary, Soft Filter Pruning (SFP) simply zeroizes pruned filters, keeping updating them in the following training epochs, thus maintaining the capacity of the network. However, SFP, together with its variants, converges much slower than HFP due to its larger search space. Our question is whether SFP-based methods and HFP can be combined to achieve better performance and speed up convergence. Firstly, we generalize SFP-based methods and HFP to analyze their characteristics. Then we propose a Gradually Hard Filter Pruning (GHFP) method to smoothly switch from SFP-based methods to HFP during training and pruning, thus maintaining a large search space at first, gradually reducing the capacity of the model to ensure a moderate convergence speed. Experimental results on CIFAR-10/100 show that our method achieves the state-of-the-art performance." @default.
- W3100848797 created "2020-11-23" @default.
- W3100848797 creator A5055016576 @default.
- W3100848797 creator A5070277080 @default.
- W3100848797 creator A5085502749 @default.
- W3100848797 date "2020-11-06" @default.
- W3100848797 modified "2023-09-27" @default.
- W3100848797 title "GHFP: Gradually Hard Filter Pruning." @default.
- W3100848797 cites W1821462560 @default.
- W3100848797 cites W1996901117 @default.
- W3100848797 cites W2102605133 @default.
- W3100848797 cites W2163605009 @default.
- W3100848797 cites W2194775991 @default.
- W3100848797 cites W2267635276 @default.
- W3100848797 cites W2302255633 @default.
- W3100848797 cites W2515385951 @default.
- W3100848797 cites W2604998962 @default.
- W3100848797 cites W2612445135 @default.
- W3100848797 cites W2764043458 @default.
- W3100848797 cites W2788653909 @default.
- W3100848797 cites W2789135445 @default.
- W3100848797 cites W2808168148 @default.
- W3100848797 cites W2921438861 @default.
- W3100848797 cites W2928560789 @default.
- W3100848797 cites W2951569836 @default.
- W3100848797 cites W2962835968 @default.
- W3100848797 cites W2962851801 @default.
- W3100848797 cites W2963363373 @default.
- W3100848797 cites W2963813662 @default.
- W3100848797 cites W2964152344 @default.
- W3100848797 cites W2964233199 @default.
- W3100848797 cites W2964266063 @default.
- W3100848797 cites W2965174861 @default.
- W3100848797 cites W2970958999 @default.
- W3100848797 cites W2997471409 @default.
- W3100848797 cites W3118608800 @default.
- W3100848797 cites W3161108526 @default.
- W3100848797 hasPublicationYear "2020" @default.
- W3100848797 type Work @default.
- W3100848797 sameAs 3100848797 @default.
- W3100848797 citedByCount "0" @default.
- W3100848797 crossrefType "posted-content" @default.
- W3100848797 hasAuthorship W3100848797A5055016576 @default.
- W3100848797 hasAuthorship W3100848797A5070277080 @default.
- W3100848797 hasAuthorship W3100848797A5085502749 @default.
- W3100848797 hasConcept C106131492 @default.
- W3100848797 hasConcept C108010975 @default.
- W3100848797 hasConcept C11413529 @default.
- W3100848797 hasConcept C119857082 @default.
- W3100848797 hasConcept C13107197 @default.
- W3100848797 hasConcept C154945302 @default.
- W3100848797 hasConcept C162324750 @default.
- W3100848797 hasConcept C173608175 @default.
- W3100848797 hasConcept C22597639 @default.
- W3100848797 hasConcept C2777303404 @default.
- W3100848797 hasConcept C31972630 @default.
- W3100848797 hasConcept C41008148 @default.
- W3100848797 hasConcept C45374587 @default.
- W3100848797 hasConcept C50522688 @default.
- W3100848797 hasConcept C50644808 @default.
- W3100848797 hasConcept C6557445 @default.
- W3100848797 hasConcept C68339613 @default.
- W3100848797 hasConcept C86803240 @default.
- W3100848797 hasConceptScore W3100848797C106131492 @default.
- W3100848797 hasConceptScore W3100848797C108010975 @default.
- W3100848797 hasConceptScore W3100848797C11413529 @default.
- W3100848797 hasConceptScore W3100848797C119857082 @default.
- W3100848797 hasConceptScore W3100848797C13107197 @default.
- W3100848797 hasConceptScore W3100848797C154945302 @default.
- W3100848797 hasConceptScore W3100848797C162324750 @default.
- W3100848797 hasConceptScore W3100848797C173608175 @default.
- W3100848797 hasConceptScore W3100848797C22597639 @default.
- W3100848797 hasConceptScore W3100848797C2777303404 @default.
- W3100848797 hasConceptScore W3100848797C31972630 @default.
- W3100848797 hasConceptScore W3100848797C41008148 @default.
- W3100848797 hasConceptScore W3100848797C45374587 @default.
- W3100848797 hasConceptScore W3100848797C50522688 @default.
- W3100848797 hasConceptScore W3100848797C50644808 @default.
- W3100848797 hasConceptScore W3100848797C6557445 @default.
- W3100848797 hasConceptScore W3100848797C68339613 @default.
- W3100848797 hasConceptScore W3100848797C86803240 @default.
- W3100848797 hasLocation W31008487971 @default.
- W3100848797 hasOpenAccess W3100848797 @default.
- W3100848797 hasPrimaryLocation W31008487971 @default.
- W3100848797 hasRelatedWork W2737100304 @default.
- W3100848797 hasRelatedWork W2897824775 @default.
- W3100848797 hasRelatedWork W2902182759 @default.
- W3100848797 hasRelatedWork W2965955046 @default.
- W3100848797 hasRelatedWork W2970632972 @default.
- W3100848797 hasRelatedWork W2995046500 @default.
- W3100848797 hasRelatedWork W3005963186 @default.
- W3100848797 hasRelatedWork W3091254925 @default.
- W3100848797 hasRelatedWork W3108068338 @default.
- W3100848797 hasRelatedWork W3109037380 @default.
- W3100848797 hasRelatedWork W3115761373 @default.
- W3100848797 hasRelatedWork W3160783538 @default.
- W3100848797 hasRelatedWork W3164847160 @default.
- W3100848797 hasRelatedWork W3187266494 @default.
- W3100848797 hasRelatedWork W3193647666 @default.
- W3100848797 hasRelatedWork W3197011103 @default.