AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

ERM-KTP: Knowledge-Level Machine Unlearning via Knowledge Transfer

Shen Lin, Xiaoyu Zhang, Chenyang Chen, Xiaofeng Chen, Willy Susilo

CVPR 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 39 citations (OpenAlex)

Read the paper

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 the ERM structure to the CNNCNN classifiers · CIFAR10, CIFAR100, and Tiny-ImageNet validationaccuracy: 89.6% -> 92.6%; 76.2% -> 77.6%; 49.3% -> 53.4%Improved
Use ERM-KTP instead of retraining from scratchResNet-20 · CIFAR10, 40% classes removedunlearning time: 668s -> 124sImproved
Use ERM-KTP for class-specific unlearningResNet-20 · CIFAR10, 40% classes removedtarget-class accuracy: 0.0%Improved
Use ERM-KTP for class-specific unlearningResNeXt-50 · CIFAR100, 40% classes removedtarget-class accuracy: 0.0%Improved
Use ERM-KTP for class-specific unlearningResNet-50 · Tiny-ImageNet, 40% classes removedtarget-class accuracy: 0.0%Improved
Use ERM-KTP instead of Fine-tuneResNeXt-50 · CIFAR100, 40% classes removedvalidation accuracy: 83.4% -> 79.4%Hurt
Use ERM-KTP instead of RfSResNet-50 · Tiny-ImageNet, 40% classes removedvalidation accuracy: 45.1% -> 53.8%Improved
Use ERM-KTP instead of RfSResNet-20 · CIFAR10, 40% classes removedvalidation accuracy: 96.3% -> 96.1%No clear effect
Use both L1-norm and inner-product regularization terms (ablation)ERM-CNN · CIFAR100CSIImproved
Remove the inner-product regularization term (ablation)ERM-CNN · CIFAR100CSIHurt
Remove the L1-norm regularization term (ablation)ERM-CNN · CIFAR100CSIHurt
Replace ERM-KTP with Fine-tune or random labelsunlearning methods · CIFAR10, ResNet-20 relearning experimentrelearning time: about the 50th epoch; around the 100th epochHurt

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