Matches in SemOpenAlex for { <https://semopenalex.org/work/W3034187113> ?p ?o ?g. }
Showing items 1 to 83 of
83
with 100 items per page.
- W3034187113 abstract "Model compression is a critical technique for cumbersome models to reduce memory consumption and accelerate inference. Here, we propose a novel method, called Variational Bayesian Sparsification, for distilling large models into small and sparse models while maintaining accuracy. Different from prior work, our approach innovatively embeds Bayesian sparsification into distillation. The core contributions are in two folds. First, the minibatch re-weighting method is proposed to dynamically balance the hard and soft knowledge, which can largely boost distillation accuracy. Then, the Bayesian deep sparse method presents to leverage the group sparseness and element sparseness simutaneously to reduce the parameter redundancy of student networks. We validate our method in MNIST, CIFAR10 and CIFAR100 datasets. Our method achieves 98.86% compression ratio with minor accuracy loss in MNIST. It is also evaluated through a compact model with only 1.99M weights on CIFAR10 and performed favorably against state-of-the-art compression methods on CFAR100, to verify the algorithm's generality." @default.
- W3034187113 created "2020-06-19" @default.
- W3034187113 creator A5031351839 @default.
- W3034187113 creator A5089582342 @default.
- W3034187113 date "2020-07-01" @default.
- W3034187113 modified "2023-09-27" @default.
- W3034187113 title "Variational Bayesian Sparsification for Distillation Compression" @default.
- W3034187113 cites W1997628954 @default.
- W3034187113 cites W2047229728 @default.
- W3034187113 cites W2164129707 @default.
- W3034187113 cites W2739879705 @default.
- W3034187113 cites W2886756692 @default.
- W3034187113 cites W2901505625 @default.
- W3034187113 cites W2963656031 @default.
- W3034187113 cites W2965837769 @default.
- W3034187113 cites W2979303251 @default.
- W3034187113 cites W4240805545 @default.
- W3034187113 doi "https://doi.org/10.1109/icme46284.2020.9102841" @default.
- W3034187113 hasPublicationYear "2020" @default.
- W3034187113 type Work @default.
- W3034187113 sameAs 3034187113 @default.
- W3034187113 citedByCount "1" @default.
- W3034187113 countsByYear W30341871132023 @default.
- W3034187113 crossrefType "proceedings-article" @default.
- W3034187113 hasAuthorship W3034187113A5031351839 @default.
- W3034187113 hasAuthorship W3034187113A5089582342 @default.
- W3034187113 hasConcept C107673813 @default.
- W3034187113 hasConcept C111919701 @default.
- W3034187113 hasConcept C11413529 @default.
- W3034187113 hasConcept C119857082 @default.
- W3034187113 hasConcept C126838900 @default.
- W3034187113 hasConcept C152124472 @default.
- W3034187113 hasConcept C153083717 @default.
- W3034187113 hasConcept C153180895 @default.
- W3034187113 hasConcept C154945302 @default.
- W3034187113 hasConcept C155846161 @default.
- W3034187113 hasConcept C159985019 @default.
- W3034187113 hasConcept C160234255 @default.
- W3034187113 hasConcept C180016635 @default.
- W3034187113 hasConcept C183115368 @default.
- W3034187113 hasConcept C190502265 @default.
- W3034187113 hasConcept C192562407 @default.
- W3034187113 hasConcept C2776214188 @default.
- W3034187113 hasConcept C41008148 @default.
- W3034187113 hasConcept C50644808 @default.
- W3034187113 hasConcept C71924100 @default.
- W3034187113 hasConceptScore W3034187113C107673813 @default.
- W3034187113 hasConceptScore W3034187113C111919701 @default.
- W3034187113 hasConceptScore W3034187113C11413529 @default.
- W3034187113 hasConceptScore W3034187113C119857082 @default.
- W3034187113 hasConceptScore W3034187113C126838900 @default.
- W3034187113 hasConceptScore W3034187113C152124472 @default.
- W3034187113 hasConceptScore W3034187113C153083717 @default.
- W3034187113 hasConceptScore W3034187113C153180895 @default.
- W3034187113 hasConceptScore W3034187113C154945302 @default.
- W3034187113 hasConceptScore W3034187113C155846161 @default.
- W3034187113 hasConceptScore W3034187113C159985019 @default.
- W3034187113 hasConceptScore W3034187113C160234255 @default.
- W3034187113 hasConceptScore W3034187113C180016635 @default.
- W3034187113 hasConceptScore W3034187113C183115368 @default.
- W3034187113 hasConceptScore W3034187113C190502265 @default.
- W3034187113 hasConceptScore W3034187113C192562407 @default.
- W3034187113 hasConceptScore W3034187113C2776214188 @default.
- W3034187113 hasConceptScore W3034187113C41008148 @default.
- W3034187113 hasConceptScore W3034187113C50644808 @default.
- W3034187113 hasConceptScore W3034187113C71924100 @default.
- W3034187113 hasLocation W30341871131 @default.
- W3034187113 hasOpenAccess W3034187113 @default.
- W3034187113 hasPrimaryLocation W30341871131 @default.
- W3034187113 hasRelatedWork W1974719870 @default.
- W3034187113 hasRelatedWork W2114746372 @default.
- W3034187113 hasRelatedWork W2144751992 @default.
- W3034187113 hasRelatedWork W2183617661 @default.
- W3034187113 hasRelatedWork W2338752163 @default.
- W3034187113 hasRelatedWork W2511279186 @default.
- W3034187113 hasRelatedWork W2532368719 @default.
- W3034187113 hasRelatedWork W2810865670 @default.
- W3034187113 hasRelatedWork W3029748970 @default.
- W3034187113 hasRelatedWork W4319453517 @default.
- W3034187113 isParatext "false" @default.
- W3034187113 isRetracted "false" @default.
- W3034187113 magId "3034187113" @default.
- W3034187113 workType "article" @default.