AI Paper Atlas

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

Is Retain Set All You Need in Machine Unlearning? Restoring Performance of Unlearned Models with Out-Of-Distribution Images

Jacopo Bonato, Marco Cotogni, Luigi Sabetta

ECCV 2024 · Label-efficient and robust learning › Knowledge distillation · approach: Few-shot and metric distillation · 17 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 distillation-trick on out-of-distribution surrogate images to the Mahalanobis forget loss (ablation)SCAR · CIFAR100 class removalAUS: 0.871 -> 0.935Improved
add Mahalanobis metric-learning forget loss to distillation-trick (ablation)SCAR · CIFAR100 class removalAUS: 0.675 -> 0.935Improved
add distillation-trick in homogeneous removal (ablation)SCAR · CIFAR100 homogeneous removalAUS: 0.409 -> 0.934Improved
use Mahalanobis distance instead of L2 distance in forget loss (ablation)SCAR · TinyImageNet class removalAUS: 0.881 -> 0.940Improved
use Mahalanobis distance instead of cosine similarity (ablation)SCAR · CIFAR100 class removalAUS: 0.933 -> 0.935No clear effect
use Gaussian noise as surrogate dataset instead of ImageNet subset (ablation)SCAR · CIFAR100 class removalAUS: 0.935 -> 0.388Hurt
use COCO subset instead of ImageNet subset as surrogate (ablation)SCAR · CIFAR100 class removalAUS: 0.935 -> 0.933No clear effect
apply SCAR to ViT-B16 instead of ResNet18SCAR · CIFAR100 class removalAUS: 0.935 -> 0.945No clear effect
use self-forget variant without forget setSCAR · CIFAR100 class removalAUS: 0.935 -> 0.929No clear effect
SCAR versus retain-set methods in homogeneous removalSCAR · CIFAR100 homogeneous removalAUS: 0.934 vs 0.965 (DUCK)Hurt

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