Matches in SemOpenAlex for { <https://semopenalex.org/work/W4280497917> ?p ?o ?g. }
Showing items 1 to 86 of
86
with 100 items per page.
- W4280497917 abstract "Deep Neural Networks (DNNs) become a practical machine learning algorithm running on various Neural Processing Units (NPUs). For higher performance and lower hardware overheads, DNN datatype reduction through quantization is proposed. Moreover, to solve the memory bottleneck caused by large data size in DNNs, several zero value-aware compression algorithms are used. However, these compression algorithms do not compress modern quantized DNNs well because of decreased zero values. We find that the latest quantized DNNs have data redundancy due to frequent narrow-width values. Because low-precision quantization reduces DNN datatypes to a simple datatype with less bits, scattered DNN data are gathered to a small number of discrete values and incur a biased data distribution. Narrow-width values occupy a large proportion of the biased distribution. Moreover, an appropriate zero run-length bits can be dynamically changed according to DNN sparsity. Based on this observation, we propose a compression algorithm that exploits narrow-width values and variable zero run-length for quantized DNNs. In experiments with three quantized DNNs, our proposed scheme yields an average compression ratio of 2.99." @default.
- W4280497917 created "2022-05-22" @default.
- W4280497917 creator A5022501553 @default.
- W4280497917 creator A5032093850 @default.
- W4280497917 creator A5036998189 @default.
- W4280497917 creator A5080268101 @default.
- W4280497917 date "2022-03-14" @default.
- W4280497917 modified "2023-10-18" @default.
- W4280497917 title "ENCORE Compression: Exploiting Narrow-width Values for Quantized Deep Neural Networks" @default.
- W4280497917 cites W2019632783 @default.
- W4280497917 cites W2040832575 @default.
- W4280497917 cites W2285660444 @default.
- W4280497917 cites W2289252105 @default.
- W4280497917 cites W2563860341 @default.
- W4280497917 cites W2613543507 @default.
- W4280497917 cites W2883920103 @default.
- W4280497917 cites W2896617691 @default.
- W4280497917 cites W2945146780 @default.
- W4280497917 cites W3043552335 @default.
- W4280497917 cites W3102633758 @default.
- W4280497917 cites W3118927099 @default.
- W4280497917 cites W4280583158 @default.
- W4280497917 doi "https://doi.org/10.23919/date54114.2022.9774545" @default.
- W4280497917 hasPublicationYear "2022" @default.
- W4280497917 type Work @default.
- W4280497917 citedByCount "3" @default.
- W4280497917 countsByYear W42804979172022 @default.
- W4280497917 crossrefType "proceedings-article" @default.
- W4280497917 hasAuthorship W4280497917A5022501553 @default.
- W4280497917 hasAuthorship W4280497917A5032093850 @default.
- W4280497917 hasAuthorship W4280497917A5036998189 @default.
- W4280497917 hasAuthorship W4280497917A5080268101 @default.
- W4280497917 hasConcept C111919701 @default.
- W4280497917 hasConcept C11413529 @default.
- W4280497917 hasConcept C115961682 @default.
- W4280497917 hasConcept C121332964 @default.
- W4280497917 hasConcept C13481523 @default.
- W4280497917 hasConcept C149635348 @default.
- W4280497917 hasConcept C152124472 @default.
- W4280497917 hasConcept C154945302 @default.
- W4280497917 hasConcept C25797200 @default.
- W4280497917 hasConcept C2780513914 @default.
- W4280497917 hasConcept C28855332 @default.
- W4280497917 hasConcept C2984842247 @default.
- W4280497917 hasConcept C41008148 @default.
- W4280497917 hasConcept C50644808 @default.
- W4280497917 hasConcept C511840579 @default.
- W4280497917 hasConcept C78548338 @default.
- W4280497917 hasConcept C9417928 @default.
- W4280497917 hasConcept C94835093 @default.
- W4280497917 hasConcept C97355855 @default.
- W4280497917 hasConceptScore W4280497917C111919701 @default.
- W4280497917 hasConceptScore W4280497917C11413529 @default.
- W4280497917 hasConceptScore W4280497917C115961682 @default.
- W4280497917 hasConceptScore W4280497917C121332964 @default.
- W4280497917 hasConceptScore W4280497917C13481523 @default.
- W4280497917 hasConceptScore W4280497917C149635348 @default.
- W4280497917 hasConceptScore W4280497917C152124472 @default.
- W4280497917 hasConceptScore W4280497917C154945302 @default.
- W4280497917 hasConceptScore W4280497917C25797200 @default.
- W4280497917 hasConceptScore W4280497917C2780513914 @default.
- W4280497917 hasConceptScore W4280497917C28855332 @default.
- W4280497917 hasConceptScore W4280497917C2984842247 @default.
- W4280497917 hasConceptScore W4280497917C41008148 @default.
- W4280497917 hasConceptScore W4280497917C50644808 @default.
- W4280497917 hasConceptScore W4280497917C511840579 @default.
- W4280497917 hasConceptScore W4280497917C78548338 @default.
- W4280497917 hasConceptScore W4280497917C9417928 @default.
- W4280497917 hasConceptScore W4280497917C94835093 @default.
- W4280497917 hasConceptScore W4280497917C97355855 @default.
- W4280497917 hasLocation W42804979171 @default.
- W4280497917 hasOpenAccess W4280497917 @default.
- W4280497917 hasPrimaryLocation W42804979171 @default.
- W4280497917 hasRelatedWork W2022517628 @default.
- W4280497917 hasRelatedWork W2054148956 @default.
- W4280497917 hasRelatedWork W2358942947 @default.
- W4280497917 hasRelatedWork W2369268110 @default.
- W4280497917 hasRelatedWork W2386783088 @default.
- W4280497917 hasRelatedWork W2887051168 @default.
- W4280497917 hasRelatedWork W3091803058 @default.
- W4280497917 hasRelatedWork W4214538768 @default.
- W4280497917 hasRelatedWork W4220868150 @default.
- W4280497917 hasRelatedWork W1999234249 @default.
- W4280497917 isParatext "false" @default.
- W4280497917 isRetracted "false" @default.
- W4280497917 workType "article" @default.