Matches in SemOpenAlex for { <https://semopenalex.org/work/W3203332826> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W3203332826 abstract "Bayesian Neural Networks (BNNs) that possess a property of uncertainty estimation have been increasingly adopted in a wide range of safety-critical AI applications which demand reliable and robust decision making, e.g., self-driving, rescue robots, medical image diagnosis. The training procedure of a probabilistic BNN model involves training an ensemble of sampled DNN models, which induces orders of magnitude larger volume of data movement than training a single DNN model. In this paper, we reveal that the root cause for BNN training inefficiency originates from the massive off-chip data transfer by Gaussian Random Variables (GRVs). To tackle this challenge, we propose a novel design that eliminates all the off-chip data transfer by GRVs through the reversed shifting of Linear Feedback Shift Registers (LFSRs) without incurring any training accuracy loss. To efficiently support our LFSR reversion strategy at the hardware level, we explore the design space of the current DNN accelerators and identify the optimal computation mapping scheme to best accommodate our strategy. By leveraging this finding, we design and prototype the first highly efficient BNN training accelerator, named Shift-BNN, that is low-cost and scalable. Extensive evaluation on five representative BNN models demonstrates that Shift-BNN achieves an average of 4.9x (up to 10.8x) boost in energy efficiency and 1.6x (up to 2.8x) speedup over the baseline DNN training accelerator." @default.
- W3203332826 created "2021-10-11" @default.
- W3203332826 creator A5005527086 @default.
- W3203332826 creator A5010481801 @default.
- W3203332826 creator A5016298390 @default.
- W3203332826 creator A5024150524 @default.
- W3203332826 creator A5043209884 @default.
- W3203332826 creator A5056956005 @default.
- W3203332826 date "2021-10-17" @default.
- W3203332826 modified "2023-09-30" @default.
- W3203332826 title "Shift-BNN: Highly-Efficient Probabilistic Bayesian Neural Network Training via Memory-Friendly Pattern Retrieving" @default.
- W3203332826 cites W1744759976 @default.
- W3203332826 cites W1983406208 @default.
- W3203332826 cites W1990315422 @default.
- W3203332826 cites W1999085092 @default.
- W3203332826 cites W2007339694 @default.
- W3203332826 cites W2066708466 @default.
- W3203332826 cites W2067523571 @default.
- W3203332826 cites W2102074146 @default.
- W3203332826 cites W2108598243 @default.
- W3203332826 cites W2194775991 @default.
- W3203332826 cites W2249651582 @default.
- W3203332826 cites W2442974303 @default.
- W3203332826 cites W2489529491 @default.
- W3203332826 cites W2626991402 @default.
- W3203332826 cites W2782718222 @default.
- W3203332826 cites W2949866178 @default.
- W3203332826 cites W2963358710 @default.
- W3203332826 cites W2979515806 @default.
- W3203332826 cites W2979754840 @default.
- W3203332826 cites W2999599167 @default.
- W3203332826 cites W3016542674 @default.
- W3203332826 cites W3043619075 @default.
- W3203332826 cites W3101380508 @default.
- W3203332826 cites W3102167048 @default.
- W3203332826 cites W3102587717 @default.
- W3203332826 cites W3105802176 @default.
- W3203332826 cites W4251999575 @default.
- W3203332826 cites W4253012315 @default.
- W3203332826 cites W4301361180 @default.
- W3203332826 doi "https://doi.org/10.1145/3466752.3480120" @default.
- W3203332826 hasPublicationYear "2021" @default.
- W3203332826 type Work @default.
- W3203332826 sameAs 3203332826 @default.
- W3203332826 citedByCount "3" @default.
- W3203332826 countsByYear W32033328262023 @default.
- W3203332826 crossrefType "proceedings-article" @default.
- W3203332826 hasAuthorship W3203332826A5005527086 @default.
- W3203332826 hasAuthorship W3203332826A5010481801 @default.
- W3203332826 hasAuthorship W3203332826A5016298390 @default.
- W3203332826 hasAuthorship W3203332826A5024150524 @default.
- W3203332826 hasAuthorship W3203332826A5043209884 @default.
- W3203332826 hasAuthorship W3203332826A5056956005 @default.
- W3203332826 hasBestOaLocation W32033328262 @default.
- W3203332826 hasConcept C119857082 @default.
- W3203332826 hasConcept C154945302 @default.
- W3203332826 hasConcept C162324750 @default.
- W3203332826 hasConcept C173608175 @default.
- W3203332826 hasConcept C175444787 @default.
- W3203332826 hasConcept C2778869765 @default.
- W3203332826 hasConcept C41008148 @default.
- W3203332826 hasConcept C48044578 @default.
- W3203332826 hasConcept C68339613 @default.
- W3203332826 hasConcept C77088390 @default.
- W3203332826 hasConceptScore W3203332826C119857082 @default.
- W3203332826 hasConceptScore W3203332826C154945302 @default.
- W3203332826 hasConceptScore W3203332826C162324750 @default.
- W3203332826 hasConceptScore W3203332826C173608175 @default.
- W3203332826 hasConceptScore W3203332826C175444787 @default.
- W3203332826 hasConceptScore W3203332826C2778869765 @default.
- W3203332826 hasConceptScore W3203332826C41008148 @default.
- W3203332826 hasConceptScore W3203332826C48044578 @default.
- W3203332826 hasConceptScore W3203332826C68339613 @default.
- W3203332826 hasConceptScore W3203332826C77088390 @default.
- W3203332826 hasFunder F4320306076 @default.
- W3203332826 hasLocation W32033328261 @default.
- W3203332826 hasLocation W32033328262 @default.
- W3203332826 hasLocation W32033328263 @default.
- W3203332826 hasOpenAccess W3203332826 @default.
- W3203332826 hasPrimaryLocation W32033328261 @default.
- W3203332826 hasRelatedWork W1589376391 @default.
- W3203332826 hasRelatedWork W2051711022 @default.
- W3203332826 hasRelatedWork W2089704382 @default.
- W3203332826 hasRelatedWork W2384714939 @default.
- W3203332826 hasRelatedWork W2499279132 @default.
- W3203332826 hasRelatedWork W271331623 @default.
- W3203332826 hasRelatedWork W2900262727 @default.
- W3203332826 hasRelatedWork W3138386522 @default.
- W3203332826 hasRelatedWork W4213121036 @default.
- W3203332826 hasRelatedWork W1506942559 @default.
- W3203332826 isParatext "false" @default.
- W3203332826 isRetracted "false" @default.
- W3203332826 magId "3203332826" @default.
- W3203332826 workType "article" @default.