Matches in SemOpenAlex for { <https://semopenalex.org/work/W4283070908> ?p ?o ?g. }
Showing items 1 to 63 of
63
with 100 items per page.
- W4283070908 abstract "Filter pruning method introduces structural sparsity by removing selected filters and is thus particularly effective for reducing complexity. Previous works empirically prune networks from the point of view that filter with smaller norm contributes less to the final results. However, such criteria has been proven sensitive to the distribution of filters, and the accuracy may hard to recover since the capacity gap is fixed once pruned. In this paper, we propose a novel filter pruning method called Asymptotic Soft Cluster Pruning (ASCP), to identify the redundancy of network based on the similarity of filters. Each filter from over-parameterized network is first distinguished by clustering, and then reconstructed to manually introduce redundancy into it. Several guidelines of clustering are proposed to better preserve feature extraction ability. After reconstruction, filters are allowed to be updated to eliminate the effect caused by mistakenly selected. Besides, various decaying strategies of the pruning rate are adopted to stabilize the pruning process and improve the final performance as well. By gradually generating more identical filters within each cluster, ASCP can remove them through channel addition operation with almost no accuracy drop. Extensive experiments on CIFAR-10 and ImageNet datasets show that our method can achieve competitive results compared with many state-of-the-art algorithms." @default.
- W4283070908 created "2022-06-19" @default.
- W4283070908 creator A5045897241 @default.
- W4283070908 creator A5081117495 @default.
- W4283070908 creator A5087011161 @default.
- W4283070908 date "2022-06-16" @default.
- W4283070908 modified "2023-09-29" @default.
- W4283070908 title "Asymptotic Soft Cluster Pruning for Deep Neural Networks" @default.
- W4283070908 doi "https://doi.org/10.48550/arxiv.2206.08186" @default.
- W4283070908 hasPublicationYear "2022" @default.
- W4283070908 type Work @default.
- W4283070908 citedByCount "0" @default.
- W4283070908 crossrefType "posted-content" @default.
- W4283070908 hasAuthorship W4283070908A5045897241 @default.
- W4283070908 hasAuthorship W4283070908A5081117495 @default.
- W4283070908 hasAuthorship W4283070908A5087011161 @default.
- W4283070908 hasBestOaLocation W42830709081 @default.
- W4283070908 hasConcept C106131492 @default.
- W4283070908 hasConcept C108010975 @default.
- W4283070908 hasConcept C111919701 @default.
- W4283070908 hasConcept C11413529 @default.
- W4283070908 hasConcept C124101348 @default.
- W4283070908 hasConcept C152124472 @default.
- W4283070908 hasConcept C153180895 @default.
- W4283070908 hasConcept C154945302 @default.
- W4283070908 hasConcept C173608175 @default.
- W4283070908 hasConcept C31972630 @default.
- W4283070908 hasConcept C3826847 @default.
- W4283070908 hasConcept C41008148 @default.
- W4283070908 hasConcept C6557445 @default.
- W4283070908 hasConcept C73555534 @default.
- W4283070908 hasConcept C86803240 @default.
- W4283070908 hasConceptScore W4283070908C106131492 @default.
- W4283070908 hasConceptScore W4283070908C108010975 @default.
- W4283070908 hasConceptScore W4283070908C111919701 @default.
- W4283070908 hasConceptScore W4283070908C11413529 @default.
- W4283070908 hasConceptScore W4283070908C124101348 @default.
- W4283070908 hasConceptScore W4283070908C152124472 @default.
- W4283070908 hasConceptScore W4283070908C153180895 @default.
- W4283070908 hasConceptScore W4283070908C154945302 @default.
- W4283070908 hasConceptScore W4283070908C173608175 @default.
- W4283070908 hasConceptScore W4283070908C31972630 @default.
- W4283070908 hasConceptScore W4283070908C3826847 @default.
- W4283070908 hasConceptScore W4283070908C41008148 @default.
- W4283070908 hasConceptScore W4283070908C6557445 @default.
- W4283070908 hasConceptScore W4283070908C73555534 @default.
- W4283070908 hasConceptScore W4283070908C86803240 @default.
- W4283070908 hasLocation W42830709081 @default.
- W4283070908 hasOpenAccess W4283070908 @default.
- W4283070908 hasPrimaryLocation W42830709081 @default.
- W4283070908 hasRelatedWork W1549839916 @default.
- W4283070908 hasRelatedWork W2154303101 @default.
- W4283070908 hasRelatedWork W2336900506 @default.
- W4283070908 hasRelatedWork W3008861477 @default.
- W4283070908 hasRelatedWork W3097702032 @default.
- W4283070908 hasRelatedWork W3119066197 @default.
- W4283070908 hasRelatedWork W3205506801 @default.
- W4283070908 hasRelatedWork W4200635345 @default.
- W4283070908 hasRelatedWork W4207000934 @default.
- W4283070908 hasRelatedWork W4226143218 @default.
- W4283070908 isParatext "false" @default.
- W4283070908 isRetracted "false" @default.
- W4283070908 workType "article" @default.