Matches in SemOpenAlex for { <https://semopenalex.org/work/W2894150216> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W2894150216 abstract "Quantization of weights and activations in Deep Neural Networks (DNNs) is a powerful technique for network compression, and has enjoyed significant attention and success. However, much of the inference-time benefit of quantization is accessible only through the use of customized hardware accelerators or by providing an FPGA implementation of quantized arithmetic. Building on prior work, we show how to construct arbitrary bit-precise signed and unsigned integer operations using a software technique which logically emph{embeds} a vector architecture with custom bit-width lanes in universally available fixed-width scalar arithmetic. We evaluate our approach on a high-end Intel Haswell processor, and an embedded ARM processor. Our approach yields very fast implementations of bit-precise custom DNN operations, which often match or exceed the performance of operations quantized to the sizes supported in native arithmetic. At the strongest level of quantization, our approach yields a maximum speedup of $thicksim6times$ on the Intel platform, and $thicksim10times$ on the ARM platform versus quantization to native 8-bit integers." @default.
- W2894150216 created "2018-10-05" @default.
- W2894150216 creator A5003800161 @default.
- W2894150216 creator A5010687192 @default.
- W2894150216 date "2019-06-01" @default.
- W2894150216 modified "2023-09-27" @default.
- W2894150216 title "Scalar Arithmetic Multiple Data: Customizable Precision for Deep Neural Networks" @default.
- W2894150216 cites W1547830536 @default.
- W2894150216 cites W1555915743 @default.
- W2894150216 cites W1558625027 @default.
- W2894150216 cites W1991716621 @default.
- W2894150216 cites W1994855795 @default.
- W2894150216 cites W2046613226 @default.
- W2894150216 cites W2094969361 @default.
- W2894150216 cites W2285660444 @default.
- W2894150216 cites W2395566064 @default.
- W2894150216 cites W2560017826 @default.
- W2894150216 cites W2585720638 @default.
- W2894150216 cites W2625457103 @default.
- W2894150216 cites W2766040906 @default.
- W2894150216 cites W2809533219 @default.
- W2894150216 cites W2962835968 @default.
- W2894150216 cites W3102169921 @default.
- W2894150216 doi "https://doi.org/10.1109/arith.2019.00018" @default.
- W2894150216 hasPublicationYear "2019" @default.
- W2894150216 type Work @default.
- W2894150216 sameAs 2894150216 @default.
- W2894150216 citedByCount "1" @default.
- W2894150216 countsByYear W28941502162020 @default.
- W2894150216 crossrefType "proceedings-article" @default.
- W2894150216 hasAuthorship W2894150216A5003800161 @default.
- W2894150216 hasAuthorship W2894150216A5010687192 @default.
- W2894150216 hasBestOaLocation W28941502162 @default.
- W2894150216 hasConcept C113775141 @default.
- W2894150216 hasConcept C11413529 @default.
- W2894150216 hasConcept C154945302 @default.
- W2894150216 hasConcept C171182647 @default.
- W2894150216 hasConcept C173608175 @default.
- W2894150216 hasConcept C2524010 @default.
- W2894150216 hasConcept C2776214188 @default.
- W2894150216 hasConcept C28855332 @default.
- W2894150216 hasConcept C33923547 @default.
- W2894150216 hasConcept C41008148 @default.
- W2894150216 hasConcept C42935608 @default.
- W2894150216 hasConcept C50644808 @default.
- W2894150216 hasConcept C57691317 @default.
- W2894150216 hasConcept C68339613 @default.
- W2894150216 hasConcept C83581934 @default.
- W2894150216 hasConcept C9390403 @default.
- W2894150216 hasConcept C94375191 @default.
- W2894150216 hasConceptScore W2894150216C113775141 @default.
- W2894150216 hasConceptScore W2894150216C11413529 @default.
- W2894150216 hasConceptScore W2894150216C154945302 @default.
- W2894150216 hasConceptScore W2894150216C171182647 @default.
- W2894150216 hasConceptScore W2894150216C173608175 @default.
- W2894150216 hasConceptScore W2894150216C2524010 @default.
- W2894150216 hasConceptScore W2894150216C2776214188 @default.
- W2894150216 hasConceptScore W2894150216C28855332 @default.
- W2894150216 hasConceptScore W2894150216C33923547 @default.
- W2894150216 hasConceptScore W2894150216C41008148 @default.
- W2894150216 hasConceptScore W2894150216C42935608 @default.
- W2894150216 hasConceptScore W2894150216C50644808 @default.
- W2894150216 hasConceptScore W2894150216C57691317 @default.
- W2894150216 hasConceptScore W2894150216C68339613 @default.
- W2894150216 hasConceptScore W2894150216C83581934 @default.
- W2894150216 hasConceptScore W2894150216C9390403 @default.
- W2894150216 hasConceptScore W2894150216C94375191 @default.
- W2894150216 hasLocation W28941502161 @default.
- W2894150216 hasLocation W28941502162 @default.
- W2894150216 hasOpenAccess W2894150216 @default.
- W2894150216 hasPrimaryLocation W28941502161 @default.
- W2894150216 hasRelatedWork W1520972637 @default.
- W2894150216 hasRelatedWork W1675009181 @default.
- W2894150216 hasRelatedWork W2050421108 @default.
- W2894150216 hasRelatedWork W2408352253 @default.
- W2894150216 hasRelatedWork W2507571256 @default.
- W2894150216 hasRelatedWork W2805002074 @default.
- W2894150216 hasRelatedWork W2897166595 @default.
- W2894150216 hasRelatedWork W2952464506 @default.
- W2894150216 hasRelatedWork W2995329031 @default.
- W2894150216 hasRelatedWork W3033586799 @default.
- W2894150216 hasRelatedWork W3089022192 @default.
- W2894150216 hasRelatedWork W3092323772 @default.
- W2894150216 hasRelatedWork W3108424616 @default.
- W2894150216 hasRelatedWork W3117297074 @default.
- W2894150216 hasRelatedWork W3118708382 @default.
- W2894150216 hasRelatedWork W3121418772 @default.
- W2894150216 hasRelatedWork W3158066599 @default.
- W2894150216 hasRelatedWork W2160596263 @default.
- W2894150216 hasRelatedWork W2187757694 @default.
- W2894150216 hasRelatedWork W2802581710 @default.
- W2894150216 isParatext "false" @default.
- W2894150216 isRetracted "false" @default.
- W2894150216 magId "2894150216" @default.
- W2894150216 workType "article" @default.