Matches in SemOpenAlex for { <https://semopenalex.org/work/W2891335318> ?p ?o ?g. }
Showing items 1 to 88 of
88
with 100 items per page.
- W2891335318 abstract "Deploying DNNs on embedded devices is a challenge because of their high memory and computational requirements. Performing DNN inference in lesser bit-width fixed point arithmetic is seen as a crucial step in realizing DNNs on embedded devices. State-of-the-art methods achieve floating point accuracy using re-training and complex activation normalization methods. In this paper we propose an accurate and efficient end-to-end DNN inference on 16-bit fixed point arithmetic. We prove that floating point accuracy can be achieved with a simple quantization method of using powers of 2 as scale factors coupled with our optimal bit-width estimation algorithm without using re-training. Additionally, it leads to efficient activation normalization using only arithmetic shifts. We show that the combination of our quantization method and activation normalization maximizes SIMD throughput resulting in 2x to 6x gain in execution time compared to floating point inference. Experimental results demonstrate that our method generalizes to different networks giving same or better accuracy compared to floating point for classification, regression and recurrent networks." @default.
- W2891335318 created "2018-09-27" @default.
- W2891335318 creator A5031971317 @default.
- W2891335318 creator A5038794404 @default.
- W2891335318 creator A5056953476 @default.
- W2891335318 creator A5070988453 @default.
- W2891335318 creator A5075033848 @default.
- W2891335318 creator A5085051959 @default.
- W2891335318 date "2018-10-01" @default.
- W2891335318 modified "2023-09-25" @default.
- W2891335318 title "Accurate and Efficient Fixed Point Inference for Deep Neural Networks" @default.
- W2891335318 cites W1605005685 @default.
- W2891335318 cites W1947481528 @default.
- W2891335318 cites W2155893237 @default.
- W2891335318 cites W2163605009 @default.
- W2891335318 cites W2185907834 @default.
- W2891335318 cites W2264905057 @default.
- W2891335318 cites W2285660444 @default.
- W2891335318 cites W2395566064 @default.
- W2891335318 cites W2962735857 @default.
- W2891335318 cites W2963114950 @default.
- W2891335318 doi "https://doi.org/10.1109/icip.2018.8451268" @default.
- W2891335318 hasPublicationYear "2018" @default.
- W2891335318 type Work @default.
- W2891335318 sameAs 2891335318 @default.
- W2891335318 citedByCount "6" @default.
- W2891335318 countsByYear W28913353182019 @default.
- W2891335318 countsByYear W28913353182020 @default.
- W2891335318 countsByYear W28913353182021 @default.
- W2891335318 countsByYear W28913353182023 @default.
- W2891335318 crossrefType "proceedings-article" @default.
- W2891335318 hasAuthorship W2891335318A5031971317 @default.
- W2891335318 hasAuthorship W2891335318A5038794404 @default.
- W2891335318 hasAuthorship W2891335318A5056953476 @default.
- W2891335318 hasAuthorship W2891335318A5070988453 @default.
- W2891335318 hasAuthorship W2891335318A5075033848 @default.
- W2891335318 hasAuthorship W2891335318A5085051959 @default.
- W2891335318 hasConcept C11413529 @default.
- W2891335318 hasConcept C134306372 @default.
- W2891335318 hasConcept C136886441 @default.
- W2891335318 hasConcept C144024400 @default.
- W2891335318 hasConcept C150552126 @default.
- W2891335318 hasConcept C154945302 @default.
- W2891335318 hasConcept C163973906 @default.
- W2891335318 hasConcept C173608175 @default.
- W2891335318 hasConcept C19165224 @default.
- W2891335318 hasConcept C2776214188 @default.
- W2891335318 hasConcept C28855332 @default.
- W2891335318 hasConcept C2984842247 @default.
- W2891335318 hasConcept C33923547 @default.
- W2891335318 hasConcept C41008148 @default.
- W2891335318 hasConcept C50644808 @default.
- W2891335318 hasConcept C61445026 @default.
- W2891335318 hasConcept C84211073 @default.
- W2891335318 hasConceptScore W2891335318C11413529 @default.
- W2891335318 hasConceptScore W2891335318C134306372 @default.
- W2891335318 hasConceptScore W2891335318C136886441 @default.
- W2891335318 hasConceptScore W2891335318C144024400 @default.
- W2891335318 hasConceptScore W2891335318C150552126 @default.
- W2891335318 hasConceptScore W2891335318C154945302 @default.
- W2891335318 hasConceptScore W2891335318C163973906 @default.
- W2891335318 hasConceptScore W2891335318C173608175 @default.
- W2891335318 hasConceptScore W2891335318C19165224 @default.
- W2891335318 hasConceptScore W2891335318C2776214188 @default.
- W2891335318 hasConceptScore W2891335318C28855332 @default.
- W2891335318 hasConceptScore W2891335318C2984842247 @default.
- W2891335318 hasConceptScore W2891335318C33923547 @default.
- W2891335318 hasConceptScore W2891335318C41008148 @default.
- W2891335318 hasConceptScore W2891335318C50644808 @default.
- W2891335318 hasConceptScore W2891335318C61445026 @default.
- W2891335318 hasConceptScore W2891335318C84211073 @default.
- W2891335318 hasLocation W28913353181 @default.
- W2891335318 hasOpenAccess W2891335318 @default.
- W2891335318 hasPrimaryLocation W28913353181 @default.
- W2891335318 hasRelatedWork W1482661175 @default.
- W2891335318 hasRelatedWork W1569442065 @default.
- W2891335318 hasRelatedWork W1793682442 @default.
- W2891335318 hasRelatedWork W2102577205 @default.
- W2891335318 hasRelatedWork W2803195680 @default.
- W2891335318 hasRelatedWork W2803935332 @default.
- W2891335318 hasRelatedWork W2891335318 @default.
- W2891335318 hasRelatedWork W2963974650 @default.
- W2891335318 hasRelatedWork W2971667683 @default.
- W2891335318 hasRelatedWork W3107622308 @default.
- W2891335318 isParatext "false" @default.
- W2891335318 isRetracted "false" @default.
- W2891335318 magId "2891335318" @default.
- W2891335318 workType "article" @default.