Matches in SemOpenAlex for { <https://semopenalex.org/work/W3024366773> ?p ?o ?g. }
- W3024366773 abstract "We introduce Bayesian Bits, a practical method for joint mixed precision quantization and pruning through gradient based optimization. Bayesian Bits employs a novel decomposition of the quantization operation, which sequentially considers doubling the bit width. At each new bit width, the residual error between the full precision value and the previously rounded value is quantized. We then decide whether or not to add this quantized residual error for a higher effective bit width and lower quantization noise. By starting with a power-of-two bit width, this decomposition will always produce hardware-friendly configurations, and through an additional 0-bit option, serves as a unified view of pruning and quantization. Bayesian Bits then introduces learnable stochastic gates, which collectively control the bit width of the given tensor. As a result, we can obtain low bit solutions by performing approximate inference over the gates, with prior distributions that encourage most of them to be switched off. We experimentally validate our proposed method on several benchmark datasets and show that we can learn pruned, mixed precision networks that provide a better trade-off between accuracy and efficiency than their static bit width equivalents." @default.
- W3024366773 created "2020-05-21" @default.
- W3024366773 creator A5005095861 @default.
- W3024366773 creator A5018863416 @default.
- W3024366773 creator A5041689212 @default.
- W3024366773 creator A5061001528 @default.
- W3024366773 creator A5070019003 @default.
- W3024366773 creator A5082671220 @default.
- W3024366773 creator A5087368991 @default.
- W3024366773 date "2020-05-14" @default.
- W3024366773 modified "2023-09-22" @default.
- W3024366773 title "Bayesian Bits: Unifying Quantization and Pruning" @default.
- W3024366773 cites W1724438581 @default.
- W3024366773 cites W1959608418 @default.
- W3024366773 cites W2047229728 @default.
- W3024366773 cites W2119717200 @default.
- W3024366773 cites W2119913432 @default.
- W3024366773 cites W2194775991 @default.
- W3024366773 cites W2242818861 @default.
- W3024366773 cites W2338908902 @default.
- W3024366773 cites W2405920868 @default.
- W3024366773 cites W2593564159 @default.
- W3024366773 cites W2619890685 @default.
- W3024366773 cites W2725061391 @default.
- W3024366773 cites W2786771851 @default.
- W3024366773 cites W2809624076 @default.
- W3024366773 cites W2810075754 @default.
- W3024366773 cites W2903260438 @default.
- W3024366773 cites W2921077359 @default.
- W3024366773 cites W2962706989 @default.
- W3024366773 cites W2962897886 @default.
- W3024366773 cites W2963122961 @default.
- W3024366773 cites W2963163009 @default.
- W3024366773 cites W2963363373 @default.
- W3024366773 cites W2963681088 @default.
- W3024366773 cites W2963828549 @default.
- W3024366773 cites W2964121744 @default.
- W3024366773 cites W2964164125 @default.
- W3024366773 cites W2964297791 @default.
- W3024366773 cites W2970971581 @default.
- W3024366773 cites W2972918064 @default.
- W3024366773 cites W2981751377 @default.
- W3024366773 cites W2982041622 @default.
- W3024366773 cites W2982479999 @default.
- W3024366773 cites W2989530497 @default.
- W3024366773 cites W2994840239 @default.
- W3024366773 cites W2995738369 @default.
- W3024366773 cites W2996548663 @default.
- W3024366773 cites W2998506323 @default.
- W3024366773 cites W3035078287 @default.
- W3024366773 hasPublicationYear "2020" @default.
- W3024366773 type Work @default.
- W3024366773 sameAs 3024366773 @default.
- W3024366773 citedByCount "11" @default.
- W3024366773 countsByYear W30243667732020 @default.
- W3024366773 countsByYear W30243667732021 @default.
- W3024366773 crossrefType "posted-content" @default.
- W3024366773 hasAuthorship W3024366773A5005095861 @default.
- W3024366773 hasAuthorship W3024366773A5018863416 @default.
- W3024366773 hasAuthorship W3024366773A5041689212 @default.
- W3024366773 hasAuthorship W3024366773A5061001528 @default.
- W3024366773 hasAuthorship W3024366773A5070019003 @default.
- W3024366773 hasAuthorship W3024366773A5082671220 @default.
- W3024366773 hasAuthorship W3024366773A5087368991 @default.
- W3024366773 hasConcept C107673813 @default.
- W3024366773 hasConcept C11413529 @default.
- W3024366773 hasConcept C13280743 @default.
- W3024366773 hasConcept C154945302 @default.
- W3024366773 hasConcept C155512373 @default.
- W3024366773 hasConcept C185798385 @default.
- W3024366773 hasConcept C205649164 @default.
- W3024366773 hasConcept C28855332 @default.
- W3024366773 hasConcept C33923547 @default.
- W3024366773 hasConcept C41008148 @default.
- W3024366773 hasConceptScore W3024366773C107673813 @default.
- W3024366773 hasConceptScore W3024366773C11413529 @default.
- W3024366773 hasConceptScore W3024366773C13280743 @default.
- W3024366773 hasConceptScore W3024366773C154945302 @default.
- W3024366773 hasConceptScore W3024366773C155512373 @default.
- W3024366773 hasConceptScore W3024366773C185798385 @default.
- W3024366773 hasConceptScore W3024366773C205649164 @default.
- W3024366773 hasConceptScore W3024366773C28855332 @default.
- W3024366773 hasConceptScore W3024366773C33923547 @default.
- W3024366773 hasConceptScore W3024366773C41008148 @default.
- W3024366773 hasLocation W30243667731 @default.
- W3024366773 hasOpenAccess W3024366773 @default.
- W3024366773 hasPrimaryLocation W30243667731 @default.
- W3024366773 hasRelatedWork W1821462560 @default.
- W3024366773 hasRelatedWork W2194775991 @default.
- W3024366773 hasRelatedWork W2242818861 @default.
- W3024366773 hasRelatedWork W2779673880 @default.
- W3024366773 hasRelatedWork W2798742790 @default.
- W3024366773 hasRelatedWork W2809624076 @default.
- W3024366773 hasRelatedWork W2884150179 @default.
- W3024366773 hasRelatedWork W2952202138 @default.
- W3024366773 hasRelatedWork W2964246969 @default.
- W3024366773 hasRelatedWork W2982479999 @default.
- W3024366773 hasRelatedWork W2998218113 @default.
- W3024366773 hasRelatedWork W3034933748 @default.
- W3024366773 hasRelatedWork W3035183452 @default.