Matches in SemOpenAlex for { <https://semopenalex.org/work/W2964133305> ?p ?o ?g. }
Showing items 1 to 79 of
79
with 100 items per page.
- W2964133305 abstract "Inference for state-of-the-art deep neural networks is computationally expensive, making them difficult to deploy on constrained hardware environments. An efficient way to reduce this complexity is to quantize the weight parameters and/or activations during training by approximating their distributions with a limited entry codebook. For very low-precisions, such as binary or ternary networks with 1-8-bit activations, the information loss from quantization leads to significant accuracy degradation due to large gradient mismatches between the forward and backward functions. In this paper, we introduce a quantization method to reduce this loss by learning a symmetric codebook for particular weight subgroups. These subgroups are determined based on their locality in the weight matrix, such that the hardware simplicity of the low-precision representations is preserved. Empirically, we show that symmetric quantization can substantially improve accuracy for networks with extremely low-precision weights and activations. We also demonstrate that this representation imposes minimal or no hardware implications to more coarse-grained approaches. Source code is available at https://www.github.com/julianfaraone/SYQ." @default.
- W2964133305 created "2019-07-30" @default.
- W2964133305 creator A5028349760 @default.
- W2964133305 creator A5031107943 @default.
- W2964133305 creator A5037482622 @default.
- W2964133305 creator A5038194948 @default.
- W2964133305 date "2018-06-01" @default.
- W2964133305 modified "2023-10-03" @default.
- W2964133305 title "SYQ: Learning Symmetric Quantization for Efficient Deep Neural Networks" @default.
- W2964133305 cites W1677182931 @default.
- W2964133305 cites W2117539524 @default.
- W2964133305 cites W2285660444 @default.
- W2964133305 cites W2583831344 @default.
- W2964133305 cites W2586654419 @default.
- W2964133305 cites W2737652487 @default.
- W2964133305 cites W2748818695 @default.
- W2964133305 doi "https://doi.org/10.1109/cvpr.2018.00452" @default.
- W2964133305 hasPublicationYear "2018" @default.
- W2964133305 type Work @default.
- W2964133305 sameAs 2964133305 @default.
- W2964133305 citedByCount "109" @default.
- W2964133305 countsByYear W29641333052018 @default.
- W2964133305 countsByYear W29641333052019 @default.
- W2964133305 countsByYear W29641333052020 @default.
- W2964133305 countsByYear W29641333052021 @default.
- W2964133305 countsByYear W29641333052022 @default.
- W2964133305 countsByYear W29641333052023 @default.
- W2964133305 crossrefType "proceedings-article" @default.
- W2964133305 hasAuthorship W2964133305A5028349760 @default.
- W2964133305 hasAuthorship W2964133305A5031107943 @default.
- W2964133305 hasAuthorship W2964133305A5037482622 @default.
- W2964133305 hasAuthorship W2964133305A5038194948 @default.
- W2964133305 hasBestOaLocation W29641333052 @default.
- W2964133305 hasConcept C108583219 @default.
- W2964133305 hasConcept C113775141 @default.
- W2964133305 hasConcept C11413529 @default.
- W2964133305 hasConcept C127759330 @default.
- W2964133305 hasConcept C154945302 @default.
- W2964133305 hasConcept C2776214188 @default.
- W2964133305 hasConcept C28855332 @default.
- W2964133305 hasConcept C2984842247 @default.
- W2964133305 hasConcept C33923547 @default.
- W2964133305 hasConcept C41008148 @default.
- W2964133305 hasConcept C48372109 @default.
- W2964133305 hasConcept C50644808 @default.
- W2964133305 hasConcept C80444323 @default.
- W2964133305 hasConcept C94375191 @default.
- W2964133305 hasConceptScore W2964133305C108583219 @default.
- W2964133305 hasConceptScore W2964133305C113775141 @default.
- W2964133305 hasConceptScore W2964133305C11413529 @default.
- W2964133305 hasConceptScore W2964133305C127759330 @default.
- W2964133305 hasConceptScore W2964133305C154945302 @default.
- W2964133305 hasConceptScore W2964133305C2776214188 @default.
- W2964133305 hasConceptScore W2964133305C28855332 @default.
- W2964133305 hasConceptScore W2964133305C2984842247 @default.
- W2964133305 hasConceptScore W2964133305C33923547 @default.
- W2964133305 hasConceptScore W2964133305C41008148 @default.
- W2964133305 hasConceptScore W2964133305C48372109 @default.
- W2964133305 hasConceptScore W2964133305C50644808 @default.
- W2964133305 hasConceptScore W2964133305C80444323 @default.
- W2964133305 hasConceptScore W2964133305C94375191 @default.
- W2964133305 hasLocation W29641333051 @default.
- W2964133305 hasLocation W29641333052 @default.
- W2964133305 hasOpenAccess W2964133305 @default.
- W2964133305 hasPrimaryLocation W29641333051 @default.
- W2964133305 hasRelatedWork W2777406049 @default.
- W2964133305 hasRelatedWork W2803935332 @default.
- W2964133305 hasRelatedWork W2971667683 @default.
- W2964133305 hasRelatedWork W3099092507 @default.
- W2964133305 hasRelatedWork W3128691579 @default.
- W2964133305 hasRelatedWork W3196579076 @default.
- W2964133305 hasRelatedWork W4214908790 @default.
- W2964133305 hasRelatedWork W4221155876 @default.
- W2964133305 hasRelatedWork W4283368760 @default.
- W2964133305 hasRelatedWork W4310823283 @default.
- W2964133305 isParatext "false" @default.
- W2964133305 isRetracted "false" @default.
- W2964133305 magId "2964133305" @default.
- W2964133305 workType "article" @default.