Matches in SemOpenAlex for { <https://semopenalex.org/work/W4328006243> ?p ?o ?g. }
Showing items 1 to 77 of
77
with 100 items per page.
- W4328006243 abstract "We propose a new structure called Expansion-Squeeze-Block by leveraging over-parameterization to train given compact neural networks. The structure expands the width of convolutional layers and adds shortcut connections for better performance without adding any nonlinearity. The expanded networks can be contracted back to the original format algebraically at inference time without loss of information. In addition, we introduce a new initialization method utilizing the weights of the pretrained original networks, which can further improve the accuracy. We evaluate our methods on CIFAR-100 and Tiny-ImageNet datasets to show their effectiveness. As evidenced by our experiments, Expansion-Squeeze-Block outperforms the baselines in most cases with average 0.81% improvements and 1.15% combined with our initialization method on CIFAR-100 and Tiny-ImageNet datasets. We also demonstrate the effectiveness of the partial expansion strategy by choosing the layers of the highest importance in VGG11-BN0.5x and ResNet20. We can save on average 58.82% #MACs and 38.82% #Params than expanding all layers." @default.
- W4328006243 created "2023-03-22" @default.
- W4328006243 creator A5003547177 @default.
- W4328006243 creator A5060963130 @default.
- W4328006243 date "2022-08-01" @default.
- W4328006243 modified "2023-09-26" @default.
- W4328006243 title "Expansion-Squeeze- Block: Linear Over-parameterization with Shortcut Connections to Train Compact Convolutional Networks" @default.
- W4328006243 cites W1677182931 @default.
- W4328006243 cites W2097117768 @default.
- W4328006243 cites W2183341477 @default.
- W4328006243 cites W2796346823 @default.
- W4328006243 cites W2963125010 @default.
- W4328006243 cites W2963163009 @default.
- W4328006243 cites W2963446712 @default.
- W4328006243 cites W2963821229 @default.
- W4328006243 cites W2963844898 @default.
- W4328006243 cites W2964081807 @default.
- W4328006243 cites W2964233199 @default.
- W4328006243 cites W2964350391 @default.
- W4328006243 cites W2965658867 @default.
- W4328006243 cites W2981609437 @default.
- W4328006243 cites W2982083293 @default.
- W4328006243 cites W3034513523 @default.
- W4328006243 cites W3109037380 @default.
- W4328006243 cites W3167976421 @default.
- W4328006243 cites W3171038842 @default.
- W4328006243 doi "https://doi.org/10.1109/bigcom57025.2022.00011" @default.
- W4328006243 hasPublicationYear "2022" @default.
- W4328006243 type Work @default.
- W4328006243 citedByCount "0" @default.
- W4328006243 crossrefType "proceedings-article" @default.
- W4328006243 hasAuthorship W4328006243A5003547177 @default.
- W4328006243 hasAuthorship W4328006243A5060963130 @default.
- W4328006243 hasConcept C11413529 @default.
- W4328006243 hasConcept C114466953 @default.
- W4328006243 hasConcept C121332964 @default.
- W4328006243 hasConcept C153180895 @default.
- W4328006243 hasConcept C154945302 @default.
- W4328006243 hasConcept C158622935 @default.
- W4328006243 hasConcept C199360897 @default.
- W4328006243 hasConcept C2524010 @default.
- W4328006243 hasConcept C2776214188 @default.
- W4328006243 hasConcept C2777210771 @default.
- W4328006243 hasConcept C33923547 @default.
- W4328006243 hasConcept C41008148 @default.
- W4328006243 hasConcept C62520636 @default.
- W4328006243 hasConcept C81363708 @default.
- W4328006243 hasConceptScore W4328006243C11413529 @default.
- W4328006243 hasConceptScore W4328006243C114466953 @default.
- W4328006243 hasConceptScore W4328006243C121332964 @default.
- W4328006243 hasConceptScore W4328006243C153180895 @default.
- W4328006243 hasConceptScore W4328006243C154945302 @default.
- W4328006243 hasConceptScore W4328006243C158622935 @default.
- W4328006243 hasConceptScore W4328006243C199360897 @default.
- W4328006243 hasConceptScore W4328006243C2524010 @default.
- W4328006243 hasConceptScore W4328006243C2776214188 @default.
- W4328006243 hasConceptScore W4328006243C2777210771 @default.
- W4328006243 hasConceptScore W4328006243C33923547 @default.
- W4328006243 hasConceptScore W4328006243C41008148 @default.
- W4328006243 hasConceptScore W4328006243C62520636 @default.
- W4328006243 hasConceptScore W4328006243C81363708 @default.
- W4328006243 hasLocation W43280062431 @default.
- W4328006243 hasOpenAccess W4328006243 @default.
- W4328006243 hasPrimaryLocation W43280062431 @default.
- W4328006243 hasRelatedWork W2175746458 @default.
- W4328006243 hasRelatedWork W2732542196 @default.
- W4328006243 hasRelatedWork W2738221750 @default.
- W4328006243 hasRelatedWork W2760085659 @default.
- W4328006243 hasRelatedWork W2883200793 @default.
- W4328006243 hasRelatedWork W2912288872 @default.
- W4328006243 hasRelatedWork W2940661641 @default.
- W4328006243 hasRelatedWork W3012978760 @default.
- W4328006243 hasRelatedWork W3052481912 @default.
- W4328006243 hasRelatedWork W3093612317 @default.
- W4328006243 isParatext "false" @default.
- W4328006243 isRetracted "false" @default.
- W4328006243 workType "article" @default.