Matches in SemOpenAlex for { <https://semopenalex.org/work/W3048791531> ?p ?o ?g. }
Showing items 1 to 98 of
98
with 100 items per page.
- W3048791531 abstract "Post-training, layer-wise quantization is preferable because it is free from retraining and is hardware-friendly. Nevertheless, accuracy degradation has occurred when a neural network model has a big difference of per-out-channel weight ranges. In particular, the MobileNet family has a tragedy drop in top-1 accuracy from 70.60% ~ 71.87% to 0.1% on the ImageNet dataset after 8-bit weight quantization. To mitigate this significant accuracy reduction, we propose a new weight equalizing shift scaler, i.e. rescaling the weight range per channel by a 4-bit binary shift, prior to a layer-wise quantization. To recover the original output range, inverse binary shifting is efficiently fused to the existing per-layer scale compounding in the fixed-computing convolutional operator of the custom neural processing unit. The binary shift is a key feature of our algorithm, which significantly improved the accuracy performance without impeding the memory footprint. As a result, our proposed method achieved a top-1 accuracy of 69.78% ~ 70.96% in MobileNets and showed robust performance in varying network models and tasks, which is competitive to channel-wise quantization results." @default.
- W3048791531 created "2020-08-18" @default.
- W3048791531 creator A5019237965 @default.
- W3048791531 creator A5037494147 @default.
- W3048791531 creator A5057696753 @default.
- W3048791531 creator A5061806967 @default.
- W3048791531 creator A5069885533 @default.
- W3048791531 date "2020-08-13" @default.
- W3048791531 modified "2023-09-26" @default.
- W3048791531 title "Weight Equalizing Shift Scaler-Coupled Post-training Quantization." @default.
- W3048791531 cites W1548556028 @default.
- W3048791531 cites W1999085092 @default.
- W3048791531 cites W2108598243 @default.
- W3048791531 cites W2117539524 @default.
- W3048791531 cites W2145607950 @default.
- W3048791531 cites W2171074980 @default.
- W3048791531 cites W2183341477 @default.
- W3048791531 cites W2194775991 @default.
- W3048791531 cites W2267635276 @default.
- W3048791531 cites W2442974303 @default.
- W3048791531 cites W2468462628 @default.
- W3048791531 cites W2503339013 @default.
- W3048791531 cites W2541839172 @default.
- W3048791531 cites W2563587242 @default.
- W3048791531 cites W2606722458 @default.
- W3048791531 cites W2612445135 @default.
- W3048791531 cites W2613718673 @default.
- W3048791531 cites W2809624076 @default.
- W3048791531 cites W2898997786 @default.
- W3048791531 cites W2915478146 @default.
- W3048791531 cites W2946948417 @default.
- W3048791531 cites W2952344559 @default.
- W3048791531 cites W2953212265 @default.
- W3048791531 cites W2963037989 @default.
- W3048791531 cites W2963114950 @default.
- W3048791531 cites W2963163009 @default.
- W3048791531 cites W2963480671 @default.
- W3048791531 cites W2963566548 @default.
- W3048791531 cites W2964299589 @default.
- W3048791531 cites W2981751377 @default.
- W3048791531 cites W2995329031 @default.
- W3048791531 cites W3008255452 @default.
- W3048791531 cites W3034940165 @default.
- W3048791531 cites W3106250896 @default.
- W3048791531 hasPublicationYear "2020" @default.
- W3048791531 type Work @default.
- W3048791531 sameAs 3048791531 @default.
- W3048791531 citedByCount "2" @default.
- W3048791531 countsByYear W30487915312021 @default.
- W3048791531 crossrefType "posted-content" @default.
- W3048791531 hasAuthorship W3048791531A5019237965 @default.
- W3048791531 hasAuthorship W3048791531A5037494147 @default.
- W3048791531 hasAuthorship W3048791531A5057696753 @default.
- W3048791531 hasAuthorship W3048791531A5061806967 @default.
- W3048791531 hasAuthorship W3048791531A5069885533 @default.
- W3048791531 hasConcept C11413529 @default.
- W3048791531 hasConcept C154945302 @default.
- W3048791531 hasConcept C28855332 @default.
- W3048791531 hasConcept C33923547 @default.
- W3048791531 hasConcept C41008148 @default.
- W3048791531 hasConcept C48372109 @default.
- W3048791531 hasConcept C81363708 @default.
- W3048791531 hasConcept C94375191 @default.
- W3048791531 hasConceptScore W3048791531C11413529 @default.
- W3048791531 hasConceptScore W3048791531C154945302 @default.
- W3048791531 hasConceptScore W3048791531C28855332 @default.
- W3048791531 hasConceptScore W3048791531C33923547 @default.
- W3048791531 hasConceptScore W3048791531C41008148 @default.
- W3048791531 hasConceptScore W3048791531C48372109 @default.
- W3048791531 hasConceptScore W3048791531C81363708 @default.
- W3048791531 hasConceptScore W3048791531C94375191 @default.
- W3048791531 hasLocation W30487915311 @default.
- W3048791531 hasOpenAccess W3048791531 @default.
- W3048791531 hasPrimaryLocation W30487915311 @default.
- W3048791531 hasRelatedWork W1505543063 @default.
- W3048791531 hasRelatedWork W1523548954 @default.
- W3048791531 hasRelatedWork W1656670690 @default.
- W3048791531 hasRelatedWork W2057251903 @default.
- W3048791531 hasRelatedWork W2057406796 @default.
- W3048791531 hasRelatedWork W2111914104 @default.
- W3048791531 hasRelatedWork W2144800312 @default.
- W3048791531 hasRelatedWork W2280341016 @default.
- W3048791531 hasRelatedWork W2353413276 @default.
- W3048791531 hasRelatedWork W2372234314 @default.
- W3048791531 hasRelatedWork W2381636312 @default.
- W3048791531 hasRelatedWork W2745133584 @default.
- W3048791531 hasRelatedWork W2758444490 @default.
- W3048791531 hasRelatedWork W2962703323 @default.
- W3048791531 hasRelatedWork W2998272665 @default.
- W3048791531 hasRelatedWork W3048812237 @default.
- W3048791531 hasRelatedWork W3173527429 @default.
- W3048791531 hasRelatedWork W3198752256 @default.
- W3048791531 hasRelatedWork W2185922140 @default.
- W3048791531 hasRelatedWork W2849724676 @default.
- W3048791531 isParatext "false" @default.
- W3048791531 isRetracted "false" @default.
- W3048791531 magId "3048791531" @default.
- W3048791531 workType "article" @default.