Matches in SemOpenAlex for { <https://semopenalex.org/work/W2896718327> ?p ?o ?g. }
Showing items 1 to 82 of
82
with 100 items per page.
- W2896718327 abstract "Unlike traditional approaches that focus on the quantization at the network level, in this work we propose to minimize the quantization effect at the tensor level. We analyze the trade-off between quantization noise and clipping distortion in low precision networks. We identify the statistics of various tensors, and derive exact expressions for the mean-square-error degradation due to clipping. By optimizing these expressions, we show marked improvements over standard quantization schemes that normally avoid clipping. For example, just by choosing the accurate clipping values, more than 40% accuracy improvement is obtained for the quantization of VGG16-BN to 4-bits of precision. Our results have many applications for the quantization of neural networks at both training and inference time. One immediate application is for a rapid deployment of neural networks to low-precision accelerators without time-consuming fine tuning or the availability of the full datasets." @default.
- W2896718327 created "2018-10-26" @default.
- W2896718327 creator A5025123281 @default.
- W2896718327 creator A5030587898 @default.
- W2896718327 creator A5057489819 @default.
- W2896718327 creator A5091675091 @default.
- W2896718327 date "2018-09-27" @default.
- W2896718327 modified "2023-09-27" @default.
- W2896718327 title "ACIQ: Analytical Clipping for Integer Quantization of neural networks" @default.
- W2896718327 cites W2043481157 @default.
- W2896718327 cites W2161160262 @default.
- W2896718327 cites W2161758346 @default.
- W2896718327 cites W2469490737 @default.
- W2896718327 cites W2753301142 @default.
- W2896718327 cites W2786771851 @default.
- W2896718327 cites W2787513570 @default.
- W2896718327 cites W2798504536 @default.
- W2896718327 cites W2798544842 @default.
- W2896718327 cites W2803739089 @default.
- W2896718327 cites W2809624076 @default.
- W2896718327 cites W2890616119 @default.
- W2896718327 cites W2954582488 @default.
- W2896718327 cites W2963122961 @default.
- W2896718327 cites W2964299589 @default.
- W2896718327 hasPublicationYear "2018" @default.
- W2896718327 type Work @default.
- W2896718327 sameAs 2896718327 @default.
- W2896718327 citedByCount "35" @default.
- W2896718327 countsByYear W28967183272019 @default.
- W2896718327 countsByYear W28967183272020 @default.
- W2896718327 countsByYear W28967183272021 @default.
- W2896718327 countsByYear W28967183272022 @default.
- W2896718327 crossrefType "posted-content" @default.
- W2896718327 hasAuthorship W2896718327A5025123281 @default.
- W2896718327 hasAuthorship W2896718327A5030587898 @default.
- W2896718327 hasAuthorship W2896718327A5057489819 @default.
- W2896718327 hasAuthorship W2896718327A5091675091 @default.
- W2896718327 hasConcept C11413529 @default.
- W2896718327 hasConcept C138885662 @default.
- W2896718327 hasConcept C154945302 @default.
- W2896718327 hasConcept C2776214188 @default.
- W2896718327 hasConcept C2776848632 @default.
- W2896718327 hasConcept C28855332 @default.
- W2896718327 hasConcept C41008148 @default.
- W2896718327 hasConcept C41895202 @default.
- W2896718327 hasConcept C50644808 @default.
- W2896718327 hasConceptScore W2896718327C11413529 @default.
- W2896718327 hasConceptScore W2896718327C138885662 @default.
- W2896718327 hasConceptScore W2896718327C154945302 @default.
- W2896718327 hasConceptScore W2896718327C2776214188 @default.
- W2896718327 hasConceptScore W2896718327C2776848632 @default.
- W2896718327 hasConceptScore W2896718327C28855332 @default.
- W2896718327 hasConceptScore W2896718327C41008148 @default.
- W2896718327 hasConceptScore W2896718327C41895202 @default.
- W2896718327 hasConceptScore W2896718327C50644808 @default.
- W2896718327 hasLocation W28967183271 @default.
- W2896718327 hasOpenAccess W2896718327 @default.
- W2896718327 hasPrimaryLocation W28967183271 @default.
- W2896718327 hasRelatedWork W1821462560 @default.
- W2896718327 hasRelatedWork W2108598243 @default.
- W2896718327 hasRelatedWork W2194775991 @default.
- W2896718327 hasRelatedWork W2267635276 @default.
- W2896718327 hasRelatedWork W2300242332 @default.
- W2896718327 hasRelatedWork W2319920447 @default.
- W2896718327 hasRelatedWork W2469490737 @default.
- W2896718327 hasRelatedWork W2524428287 @default.
- W2896718327 hasRelatedWork W2786771851 @default.
- W2896718327 hasRelatedWork W2809624076 @default.
- W2896718327 hasRelatedWork W2884150179 @default.
- W2896718327 hasRelatedWork W2912168260 @default.
- W2896718327 hasRelatedWork W2963114950 @default.
- W2896718327 hasRelatedWork W2963122961 @default.
- W2896718327 hasRelatedWork W2963163009 @default.
- W2896718327 hasRelatedWork W2964264300 @default.
- W2896718327 hasRelatedWork W2981751377 @default.
- W2896718327 hasRelatedWork W2998218113 @default.
- W2896718327 hasRelatedWork W3034940165 @default.
- W2896718327 hasRelatedWork W3118608800 @default.
- W2896718327 isParatext "false" @default.
- W2896718327 isRetracted "false" @default.
- W2896718327 magId "2896718327" @default.
- W2896718327 workType "article" @default.