AI Paper Atlas

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

Explaining Knowledge Distillation by Quantifying the Knowledge

Xu Cheng, Zhefan Rao, Yilan Chen, Quanshi Zhang

CVPR 2020 · Label-efficient and robust learning › Knowledge distillation · approach: Self and data-free distillation · 121 citations (OpenAlex)

Check 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
Distill from the teacher instead of training from raw dataVGG-16 · CUB200-2011, FC1foreground visual concepts: 22.50 -> 29.57Improved
Distill from the teacher instead of training from raw dataVGG-16 · CUB200-2011, FC1foreground ratio λ: 0.68 -> 0.75Improved
Distill from the teacher instead of training from raw dataVGG-16 · CUB200-2011, FC2foreground visual concepts: 23.31 -> 32.92Improved
Distill from the teacher instead of training from raw dataVGG-16 · CUB200-2011, FC2foreground ratio λ: 0.67 -> 0.75Improved
Distill from the teacher instead of training from raw dataVGG-16 · ILSVRC-2013 DET, FC1foreground visual concepts: 36.47 -> 35.98Hurt
Distill from the teacher instead of training from raw dataVGG-16 · ILSVRC-2013 DET, FC1foreground ratio λ: 0.81 -> 0.84Improved
Distill from the teacher instead of training from raw dataVGG-16 · ILSVRC-2013 DET, FC2foreground visual concepts: 36.35 -> 42.74Improved
Distill from the teacher instead of training from raw dataVGG-16 · ILSVRC-2013 DET, FC2foreground ratio λ: 0.82 -> 0.82No clear effect
Use knowledge distillation rather than raw-data trainingCompared student and baseline networks · CUB200-2011, ILSVRC-2013 DET, and Pascal VOC 2012learning speed and simultaneity (D_mean and D_std)Mixed
Use knowledge distillation rather than raw-data trainingCompared student and baseline networks · CUB200-2011, ILSVRC-2013 DET, and Pascal VOC 2012optimization-direction stability (ρ)Mixed

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