AI Paper Atlas

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

Knowledge Distillation via Route Constrained Optimization

Xiao Jin, Baoyun Peng, Yichao Wu, Yu Liu, Jiaheng Liu, Ding Liang, Junjie Yan, Xiaolin Hu

ICCV 2019 · Label-efficient and robust learning › Knowledge distillation · approach: Rethinking distillation objectives · 205 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 route-constrained distillation (RCO) instead of KDMobileNetV2 · CIFAR-100top-1 accuracy: 68.71 -> 70.85Improved
add RCO instead of KDMobileNetV2 · ImageNet valtop-1 accuracy: 66.75 -> 68.21Improved
add RCO instead of FitNet hint learningMobileNetV2 · MegaFace 1M distractorstop-1 accuracy: 81.02 -> 84.3Improved
use one-stage EEI with M_gap 10 instead of KD (ablation)MobileNetV2 T=4 w=0.35 · CIFAR-100top-1 accuracy: 60.79 -> 65.01Improved
use RCO M_gap 10 instead of DML (ablation)MobileNetV2 T=4 w=0.35 · CIFAR-100top-1 accuracy: 61.13 -> 65.01Improved
use M_gap 10 instead of per-iteration mimicking (ablation)RCO one-stage · CIFAR-100top-1 accuracy: 61.63 -> 65.01Improved
use M_gap 20 instead of 10 (ablation)RCO one-stage · CIFAR-100top-1 accuracy: 65.01 -> 63.88Hurt
use greedy search anchors instead of EEI-4 (ablation)RCO multi-stage · CIFAR-100top-1 accuracy: 65.27 -> 65.41Improved
use EEI-2 instead of one-stage EEI (ablation)RCO · CIFAR-100top-1 accuracy: 65.01 -> 61.43Hurt
use more anchor points: EEI-3 vs EEI-2 (ablation)RCO multi-stage · CIFAR-100top-1 accuracy: 61.43 -> 63.34Improved
use RCO on smaller studentsMobileNetV2 variants · CIFAR-100top-1 accuracyImproved
add RCO instead of KD for noise robustnessMobileNetV2 · CIFAR-100 noisy testloss gapImproved

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