AI Paper Atlas

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

Prime-Aware Adaptive Distillation

Youcai Zhang, Zhonghao Lan, Yuchen Dai, Fangao Zeng, Yan Bai, Jie Chang, Yichen Wei

ECCV 2020 · Label-efficient and robust learning › Knowledge distillation · approach: Distillation with causal and adaptive guidance · 49 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
add hard-mining weighting (large weights for hard samples) (ablation)L2 distillation · CIFAR100top-1 accuracy (%): 72.85 -> 72.12Hurt
add hard-mining weighting (ablation)L2 distillation · CUB-200R@1 (%): 58.6 -> 55.7Hurt
add hard-mining weighting (ablation)L2 distillation · TinyImageNettop-1 accuracy (%): 63.08 -> 62.57Hurt
discard the hardest samples in each batch (ablation)L2 distillation · CIFAR100top-1 accuracy (%): 72.85 -> 72.91No clear effect
discard the hardest samples in each batch (ablation)L2 distillation · Pascal VOC07mmAP: 45.53 -> 45.94Improved
add soft-exp weighting (ablation)L2 distillation · CIFAR100top-1 accuracy (%): 72.85 -> 73.54Improved
add soft-poly weighting (ablation)L2 distillation · CUB-200R@1 (%): 58.6 -> 60.4Improved
vary hyper-parameter of soft-exp weighting (0.1 to 10) (ablation)L2 distillation · TinyImageNettop-1 accuracy (%): 63.1 baseline; 61.2 at 0.1, 63.6 at 5Mixed
use PAD learned weights instead of soft-exp (ablation)L2 distillation · TinyImageNettop-1 accuracy (%): 63.64 -> 65.54Improved
add full PAD (variance + log term) over weights from PAD (ablation)L2 distillation · Pascal VOC07mmAP: 46.52 -> 46.92Improved
add warm up to L2 distillation (ablation)L2 distillation · CIFAR100top-1 accuracy (%): 72.85 -> 73.13Improved
add PAD to L2 distillationL2 distillation · CIFAR100top-1 accuracy (%): 72.85 -> 74.06Improved

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