Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385482877> ?p ?o ?g. }
Showing items 1 to 70 of
70
with 100 items per page.
- W4385482877 abstract "Vector Quantization (VQ) is an appealing model compression method to obtain a tiny model with less accuracy loss. While methods to obtain better codebooks and codes under fixed clustering dimensionality have been extensively studied, optimizations via the reduction of subvector dimensionality are not carefully considered. This paper reports our recent progress on model compression with the combination of dimensionality reduction and vector quantization, proposing Low-Rank Representation Vector Quantization (LR <sup xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>2</sup> VQ). LR <sup xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>2</sup> VQ joins low-rank representation with subvector clustering to construct a new kind of building block that is optimized by end-to-end training. In our method, the compression ratio could be directly controlled by the dimensionality of subvectors, and the final accuracy is solely determined by clustering dimensionality <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$tilde{d}$</tex> . We recognize <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$tilde{d}$</tex> as a trade-off between low-rank approximation error and clustering error and carry out both theoretical analysis and experimental observations that empower the estimation of the proper <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$tilde{d}$</tex> before fine-tuning. With a proper <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$tilde{d}$</tex> , we evaluate LR <sup xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>2</sup> VQ with ResNet-18/ResNet-50 on ImageNet classification datasets, achieving 2.8%/1.0% top-1 accuracy improvements over the current state-of-the-art model compression algorithms with 43×/31× compression factor." @default.
- W4385482877 created "2023-08-03" @default.
- W4385482877 creator A5006894022 @default.
- W4385482877 creator A5047316686 @default.
- W4385482877 creator A5056136429 @default.
- W4385482877 date "2023-06-18" @default.
- W4385482877 modified "2023-09-27" @default.
- W4385482877 title "Learning Low-Rank Representations for Model Compression" @default.
- W4385482877 cites W1677182931 @default.
- W4385482877 cites W2117539524 @default.
- W4385482877 cites W2124509324 @default.
- W4385482877 cites W2147717514 @default.
- W4385482877 cites W2194775991 @default.
- W4385482877 cites W2963125010 @default.
- W4385482877 cites W2963163009 @default.
- W4385482877 cites W2963363373 @default.
- W4385482877 cites W2964081807 @default.
- W4385482877 cites W2964233199 @default.
- W4385482877 cites W2982083293 @default.
- W4385482877 cites W3001340915 @default.
- W4385482877 doi "https://doi.org/10.1109/ijcnn54540.2023.10191936" @default.
- W4385482877 hasPublicationYear "2023" @default.
- W4385482877 type Work @default.
- W4385482877 citedByCount "0" @default.
- W4385482877 crossrefType "proceedings-article" @default.
- W4385482877 hasAuthorship W4385482877A5006894022 @default.
- W4385482877 hasAuthorship W4385482877A5047316686 @default.
- W4385482877 hasAuthorship W4385482877A5056136429 @default.
- W4385482877 hasBestOaLocation W43854828772 @default.
- W4385482877 hasConcept C111030470 @default.
- W4385482877 hasConcept C11413529 @default.
- W4385482877 hasConcept C114614502 @default.
- W4385482877 hasConcept C153180895 @default.
- W4385482877 hasConcept C154945302 @default.
- W4385482877 hasConcept C164226766 @default.
- W4385482877 hasConcept C199833920 @default.
- W4385482877 hasConcept C28855332 @default.
- W4385482877 hasConcept C33923547 @default.
- W4385482877 hasConcept C41008148 @default.
- W4385482877 hasConcept C70518039 @default.
- W4385482877 hasConcept C73555534 @default.
- W4385482877 hasConceptScore W4385482877C111030470 @default.
- W4385482877 hasConceptScore W4385482877C11413529 @default.
- W4385482877 hasConceptScore W4385482877C114614502 @default.
- W4385482877 hasConceptScore W4385482877C153180895 @default.
- W4385482877 hasConceptScore W4385482877C154945302 @default.
- W4385482877 hasConceptScore W4385482877C164226766 @default.
- W4385482877 hasConceptScore W4385482877C199833920 @default.
- W4385482877 hasConceptScore W4385482877C28855332 @default.
- W4385482877 hasConceptScore W4385482877C33923547 @default.
- W4385482877 hasConceptScore W4385482877C41008148 @default.
- W4385482877 hasConceptScore W4385482877C70518039 @default.
- W4385482877 hasConceptScore W4385482877C73555534 @default.
- W4385482877 hasLocation W43854828771 @default.
- W4385482877 hasLocation W43854828772 @default.
- W4385482877 hasOpenAccess W4385482877 @default.
- W4385482877 hasPrimaryLocation W43854828771 @default.
- W4385482877 hasRelatedWork W1520675566 @default.
- W4385482877 hasRelatedWork W1965275221 @default.
- W4385482877 hasRelatedWork W2066259560 @default.
- W4385482877 hasRelatedWork W2110350792 @default.
- W4385482877 hasRelatedWork W2139206670 @default.
- W4385482877 hasRelatedWork W2172836935 @default.
- W4385482877 hasRelatedWork W3110687914 @default.
- W4385482877 hasRelatedWork W4214747999 @default.
- W4385482877 hasRelatedWork W4312907272 @default.
- W4385482877 hasRelatedWork W2087197267 @default.
- W4385482877 isParatext "false" @default.
- W4385482877 isRetracted "false" @default.
- W4385482877 workType "article" @default.