AI Paper Atlas

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

Improving Knowledge Distillation via Regularizing Feature Direction and Norm

Yuzhu Wang, Lechao Cheng, Manni Duan, Yongheng Wang, Zunlei Feng, Shu Kong

ECCV 2024 · Label-efficient and robust learning › Knowledge distillation · approach: Distillation with causal and adaptive guidance · 22 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 feature direction regularization (cosine to teacher class-mean) to KD (ablation)KD · CIFAR-100 (R50 -> MV2)top-1 accuracy: 67.65 -> 69.18Improved
add feature norm regularization (SIFN) to KD (ablation)KD · CIFAR-100 (R50 -> MV2)top-1 accuracy: 67.65 -> 69.32Improved
use cosine rather than InfoNCE for direction regularization (ablation)KD · CIFAR-100 (R56 -> R20)top-1 accuracy: 70.73 -> 71.75Improved
naively add direction and norm regularizers together (ablation)KD · CIFAR-100 (R50 -> MV2)top-1 accuracy: 69.18 -> 68.62No clear effect
use unified dino-loss instead of separate direction plus norm terms (ablation)KD · CIFAR-100 (R50 -> MV2)top-1 accuracy: 69.07 -> 70.10Improved
use teacher class-mean features instead of classifier weights as direction targets (ablation)dino-loss · CIFAR-100 (ResNet-50 -> MobileNet-V2)top-1 accuracy: 69.76 -> 70.10Improved
match teacher classifier weights with L2 loss (ablation)KD · CIFAR-100 (ResNet-50 -> MobileNet-V2)top-1 accuracy: 67.65 -> 66.81Hurt
apply dino-loss at earlier layers instead of penultimate layer (ablation)KD++ · CIFAR-100 (ResNet-50 -> MobileNet-V2)top-1 accuracy: 70.10 -> 69.44Hurt
add dino-loss to KDKD · ImageNet val (ResNet-34 -> ResNet-18)top-1 accuracy: 70.66 -> 71.98Improved
add dino-loss to ReviewKDReviewKD · ImageNet val (ResNet-34 -> ResNet-18)top-1 accuracy: 71.62 -> 71.64No clear effect
add dino-loss to KD while scaling teacher from ResNet-34 to ResNet-152KD · ImageNet val (ResNet-18 student)top-1 accuracy: 71.99 -> 72.59Improved
scale teacher from ResNet-34 to ResNet-152 for plain KDKD · ImageNet val (ResNet-18 student)top-1 accuracy: 70.67 -> 71.13Mixed

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