AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Self-supervised Knowledge Distillation Using Singular Value Decomposition

Seung Hyun Lee, Dae Ha Kim, Byung Cheol Song

ECCV 2018 · Label-efficient and robust learning › Knowledge distillation · approach: Rethinking distillation objectives · 168 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace Yim et al. correlation knowledge with SVD+RBF distillationVGG student · CIFAR-100accuracy (%): 64.54 -> 65.05Improved
replace Yim et al. knowledge with SVD+RBF distillationMobileNet student · CIFAR-100accuracy (%): 56.53 -> 58.15Improved
replace Yim et al. knowledge with SVD+RBF distillationResNext student · CIFAR-100accuracy (%): 63.60 -> 65.43Improved
distill knowledge into a strided student with low-resolution feature mapsVGG student · CIFAR-100 (strided)accuracy (%): 54.17 -> 61.15Improved
add one-stage self-supervised multi-task training instead of two-stage (ablation)SVD+RBF distillation · CIFAR-100 (VGG)accuracy (%): 65.05 -> 65.54Improved
add one-stage multi-task training instead of two-stage (ablation)Yim et al. distillation · CIFAR-100 (VGG)accuracy (%): 64.54 -> 64.89Improved
use SVD+RBF distillation instead of Yim et al. with 10% labelsVGG semi-supervised · CIFAR-100 (10% labels)accuracy (%): 33.92 -> 49.68Improved
use SVD+RBF distillation instead of Yim et al. with 50% labelsVGG semi-supervised · CIFAR-100 (50% labels)accuracy (%): 55.21 -> 60.01Improved
increase number of DFVs from 1 to 2 (ablation)pooling student · CIFAR-100accuracy (%): 65.54 -> 66.33Improved
increase number of DFVs from 2 to 16 (ablation)pooling student · CIFAR-100accuracy (%): 66.33 -> 65.15Hurt
increase number of DFVs from 1 to 16 (ablation)strided student · CIFAR-100accuracy (%): 61.28 -> 62.00Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy