Matches in SemOpenAlex for { <https://semopenalex.org/work/W3100118801> ?p ?o ?g. }
Showing items 1 to 84 of
84
with 100 items per page.
- W3100118801 endingPage "2146" @default.
- W3100118801 startingPage "2142" @default.
- W3100118801 abstract "This brief presents a memory-efficient CNN accelerator design for resource-constrained devices in Internet of Things (IoT) and autonomous systems. A segmented logarithmic (SegLog) quantization method is exploited to mitigate the on-chip memory and bandwidth requirements, thus accommodating more processing elements (PEs) in a given chip area to organize a reconfigurable multi-cluster architecture. The evaluation results show that SegLog quantization can achieve <inline-formula xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink> <tex-math notation=LaTeX>$6.4times $ </tex-math></inline-formula> model compression with less than 2.5% accuracy loss on various CNNs. An ASIC implementation with 168 PEs configuration is validated in a 40-nm CMOS process, with 2.54 TOPs/W energy efficiency and 0.8 mm <sup xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>2</sup> chip area reported. The accelerator has also been implemented on FPGA with 1512 PEs configured and 468 kB on-chip memory, achieving a 1.29 GOPs/kB memory efficiency. Compared with the state-of-the-art accelerators, our ASIC implementation enhances area efficiency and arithmetic intensity by <inline-formula xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink> <tex-math notation=LaTeX>$1.94times $ </tex-math></inline-formula> and <inline-formula xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink> <tex-math notation=LaTeX>$5.62times $ </tex-math></inline-formula> , while the FPGA implementation achieves the memory efficiency improvement by a factor of <inline-formula xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink> <tex-math notation=LaTeX>$2.34times $ </tex-math></inline-formula> ." @default.
- W3100118801 created "2020-11-23" @default.
- W3100118801 creator A5006594466 @default.
- W3100118801 creator A5024591419 @default.
- W3100118801 creator A5041820944 @default.
- W3100118801 creator A5043039025 @default.
- W3100118801 creator A5068929319 @default.
- W3100118801 creator A5070412652 @default.
- W3100118801 creator A5071243431 @default.
- W3100118801 date "2021-06-01" @default.
- W3100118801 modified "2023-10-17" @default.
- W3100118801 title "A Memory-Efficient CNN Accelerator Using Segmented Logarithmic Quantization and Multi-Cluster Architecture" @default.
- W3100118801 cites W2588820778 @default.
- W3100118801 cites W2694935213 @default.
- W3100118801 cites W2766143712 @default.
- W3100118801 cites W2802238174 @default.
- W3100118801 cites W2807127337 @default.
- W3100118801 cites W2914552857 @default.
- W3100118801 cites W2931301255 @default.
- W3100118801 cites W2945146780 @default.
- W3100118801 cites W2949619037 @default.
- W3100118801 cites W2973613048 @default.
- W3100118801 cites W2977634443 @default.
- W3100118801 cites W2979455536 @default.
- W3100118801 cites W3009636589 @default.
- W3100118801 doi "https://doi.org/10.1109/tcsii.2020.3038897" @default.
- W3100118801 hasPublicationYear "2021" @default.
- W3100118801 type Work @default.
- W3100118801 sameAs 3100118801 @default.
- W3100118801 citedByCount "6" @default.
- W3100118801 countsByYear W31001188012021 @default.
- W3100118801 countsByYear W31001188012022 @default.
- W3100118801 countsByYear W31001188012023 @default.
- W3100118801 crossrefType "journal-article" @default.
- W3100118801 hasAuthorship W3100118801A5006594466 @default.
- W3100118801 hasAuthorship W3100118801A5024591419 @default.
- W3100118801 hasAuthorship W3100118801A5041820944 @default.
- W3100118801 hasAuthorship W3100118801A5043039025 @default.
- W3100118801 hasAuthorship W3100118801A5068929319 @default.
- W3100118801 hasAuthorship W3100118801A5070412652 @default.
- W3100118801 hasAuthorship W3100118801A5071243431 @default.
- W3100118801 hasConcept C11413529 @default.
- W3100118801 hasConcept C134306372 @default.
- W3100118801 hasConcept C165005293 @default.
- W3100118801 hasConcept C28855332 @default.
- W3100118801 hasConcept C33923547 @default.
- W3100118801 hasConcept C39927690 @default.
- W3100118801 hasConcept C41008148 @default.
- W3100118801 hasConcept C76155785 @default.
- W3100118801 hasConcept C77390884 @default.
- W3100118801 hasConcept C9390403 @default.
- W3100118801 hasConceptScore W3100118801C11413529 @default.
- W3100118801 hasConceptScore W3100118801C134306372 @default.
- W3100118801 hasConceptScore W3100118801C165005293 @default.
- W3100118801 hasConceptScore W3100118801C28855332 @default.
- W3100118801 hasConceptScore W3100118801C33923547 @default.
- W3100118801 hasConceptScore W3100118801C39927690 @default.
- W3100118801 hasConceptScore W3100118801C41008148 @default.
- W3100118801 hasConceptScore W3100118801C76155785 @default.
- W3100118801 hasConceptScore W3100118801C77390884 @default.
- W3100118801 hasConceptScore W3100118801C9390403 @default.
- W3100118801 hasFunder F4320321001 @default.
- W3100118801 hasFunder F4320329902 @default.
- W3100118801 hasIssue "6" @default.
- W3100118801 hasLocation W31001188011 @default.
- W3100118801 hasOpenAccess W3100118801 @default.
- W3100118801 hasPrimaryLocation W31001188011 @default.
- W3100118801 hasRelatedWork W161057913 @default.
- W3100118801 hasRelatedWork W1677790289 @default.
- W3100118801 hasRelatedWork W1698641079 @default.
- W3100118801 hasRelatedWork W2014526480 @default.
- W3100118801 hasRelatedWork W2064657546 @default.
- W3100118801 hasRelatedWork W2285582634 @default.
- W3100118801 hasRelatedWork W2374497491 @default.
- W3100118801 hasRelatedWork W2393944146 @default.
- W3100118801 hasRelatedWork W2596448771 @default.
- W3100118801 hasRelatedWork W4313034453 @default.
- W3100118801 hasVolume "68" @default.
- W3100118801 isParatext "false" @default.
- W3100118801 isRetracted "false" @default.
- W3100118801 magId "3100118801" @default.
- W3100118801 workType "article" @default.