Machine Unlearning on SVHN Dt (test)
90.72AccuracyCertified Hessian
Evaluation results
| Method | Links | |
|---|---|---|
| Certified HessianUnlearning Approach=Cert.-Hess., Proportion of remaining data Dr used (@)=0.352026.06 | 90.72 | |
| Fine-tuneUnlearning Approach=Fine-tune, Proportion of remaining data Dr used (@)=0.32026.06 | 90.5 | |
| Relabel with Fine-tuneUnlearning Approach=Relab.+FT, Proportion of remaining data Dr used (@)=0.352026.06 | 90.4 | |
| HessianUnlearning Approach=Hessian, Proportion of remaining data Dr used (@)=0.32026.06 | 89.98 | |
| Pre-unlearnUnlearning Approach=Pre-unlearn2026.06 | 89.94 | |
| Gradient Ascent with Fine-tuneUnlearning Approach=GA+FT, Proportion of remaining data Dr used (@)=0.32026.06 | 89.45 | |
| Adversarial RetrainUnlearning Approach=Adv. Retr., Proportion of remaining data Dr used (@)=0.32026.06 | 89.43 | |
| FisherUnlearning Approach=Fisher, Proportion of remaining data Dr used (@)=0.352026.06 | 89.18 | |
| RetrainUnlearning Approach=Retrain, Proportion of remaining data Dr used (@)=0.32026.06 | 87.29 | |
| Gradient AscentUnlearning Approach=GA, Proportion of remaining data Dr used (@)=0.052026.06 | 40.83 | |
| RelabelUnlearning Approach=Relabel, Proportion of remaining data Dr used (@)=0.052026.06 | 35.64 |