Matches in SemOpenAlex for { <https://semopenalex.org/work/W3036842007> ?p ?o ?g. }
Showing items 1 to 90 of
90
with 100 items per page.
- W3036842007 abstract "Pruning and quantization are proven methods for improving the performance and storage efficiency of convolutional neural networks (CNNs). Pruning removes near-zero weights in tensors and masks weak connections between neurons in neighbouring layers. Quantization reduces the precision of weights by replacing them with numerically similar values that require less storage. In this paper we identify another form of redundancy in CNN weight tensors, in the form of repeated patterns of similar values. We observe that pruning and quantization both tend to drastically increase the number of repeated patterns in the weight tensors. We investigate several compression schemes to take advantage of this structure in CNN weight data, including multiple forms of Huffman coding, and other approaches inspired by block sparse matrix formats. We evaluate our approach on several well-known CNNs and find that we can achieve compaction ratios of 1.4x to 3.1x in addition to the saving from pruning and quantization." @default.
- W3036842007 created "2020-06-25" @default.
- W3036842007 creator A5003800161 @default.
- W3036842007 creator A5041572550 @default.
- W3036842007 date "2021-05-04" @default.
- W3036842007 modified "2023-09-27" @default.
- W3036842007 title "Exploiting Weight Redundancy in CNNs: Beyond Pruning and Quantization" @default.
- W3036842007 cites W107968503 @default.
- W3036842007 cites W1506342804 @default.
- W3036842007 cites W1990832096 @default.
- W3036842007 cites W2106721790 @default.
- W3036842007 cites W2114766824 @default.
- W3036842007 cites W2119144962 @default.
- W3036842007 cites W2125389748 @default.
- W3036842007 cites W2161591461 @default.
- W3036842007 cites W2276892413 @default.
- W3036842007 cites W2279098554 @default.
- W3036842007 cites W2337344472 @default.
- W3036842007 cites W2483966489 @default.
- W3036842007 cites W2495425901 @default.
- W3036842007 cites W2554302513 @default.
- W3036842007 cites W2604319603 @default.
- W3036842007 cites W2657126969 @default.
- W3036842007 cites W2736953746 @default.
- W3036842007 cites W2752037867 @default.
- W3036842007 cites W2798544842 @default.
- W3036842007 cites W2962965870 @default.
- W3036842007 cites W2963363373 @default.
- W3036842007 cites W2963674932 @default.
- W3036842007 cites W2963893493 @default.
- W3036842007 cites W2963981420 @default.
- W3036842007 cites W2964233199 @default.
- W3036842007 cites W3104393472 @default.
- W3036842007 cites W587794757 @default.
- W3036842007 hasPublicationYear "2021" @default.
- W3036842007 type Work @default.
- W3036842007 sameAs 3036842007 @default.
- W3036842007 citedByCount "0" @default.
- W3036842007 crossrefType "journal-article" @default.
- W3036842007 hasAuthorship W3036842007A5003800161 @default.
- W3036842007 hasAuthorship W3036842007A5041572550 @default.
- W3036842007 hasConcept C111919701 @default.
- W3036842007 hasConcept C11413529 @default.
- W3036842007 hasConcept C152124472 @default.
- W3036842007 hasConcept C153180895 @default.
- W3036842007 hasConcept C154945302 @default.
- W3036842007 hasConcept C28855332 @default.
- W3036842007 hasConcept C33923547 @default.
- W3036842007 hasConcept C41008148 @default.
- W3036842007 hasConcept C46900642 @default.
- W3036842007 hasConcept C78548338 @default.
- W3036842007 hasConcept C81363708 @default.
- W3036842007 hasConceptScore W3036842007C111919701 @default.
- W3036842007 hasConceptScore W3036842007C11413529 @default.
- W3036842007 hasConceptScore W3036842007C152124472 @default.
- W3036842007 hasConceptScore W3036842007C153180895 @default.
- W3036842007 hasConceptScore W3036842007C154945302 @default.
- W3036842007 hasConceptScore W3036842007C28855332 @default.
- W3036842007 hasConceptScore W3036842007C33923547 @default.
- W3036842007 hasConceptScore W3036842007C41008148 @default.
- W3036842007 hasConceptScore W3036842007C46900642 @default.
- W3036842007 hasConceptScore W3036842007C78548338 @default.
- W3036842007 hasConceptScore W3036842007C81363708 @default.
- W3036842007 hasLocation W30368420071 @default.
- W3036842007 hasOpenAccess W3036842007 @default.
- W3036842007 hasPrimaryLocation W30368420071 @default.
- W3036842007 hasRelatedWork W1981114848 @default.
- W3036842007 hasRelatedWork W2130080419 @default.
- W3036842007 hasRelatedWork W2201808379 @default.
- W3036842007 hasRelatedWork W2754084392 @default.
- W3036842007 hasRelatedWork W2791795660 @default.
- W3036842007 hasRelatedWork W2904411247 @default.
- W3036842007 hasRelatedWork W2942511596 @default.
- W3036842007 hasRelatedWork W2946608096 @default.
- W3036842007 hasRelatedWork W2947963429 @default.
- W3036842007 hasRelatedWork W2949801355 @default.
- W3036842007 hasRelatedWork W2953137288 @default.
- W3036842007 hasRelatedWork W2963015611 @default.
- W3036842007 hasRelatedWork W3002829146 @default.
- W3036842007 hasRelatedWork W3014964857 @default.
- W3036842007 hasRelatedWork W3030293252 @default.
- W3036842007 hasRelatedWork W3140640946 @default.
- W3036842007 hasRelatedWork W3150160418 @default.
- W3036842007 hasRelatedWork W3199950635 @default.
- W3036842007 hasRelatedWork W580123053 @default.
- W3036842007 hasRelatedWork W2559186914 @default.
- W3036842007 isParatext "false" @default.
- W3036842007 isRetracted "false" @default.
- W3036842007 magId "3036842007" @default.
- W3036842007 workType "article" @default.