Machine Unlearning on Convex cubic loss d=1 200,000 (test)
0.0017SigmaPerturbed Newton
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Perturbed Newtonepsilon=800.02026.07 | 0.0017 | 10.1983 | 9.4159 | |
| Perturbed Newtonepsilon=800.02026.07 | 0.0017 | 10.1983 | 9.4159 | |
| Perturbed Newtonepsilon=600.02026.07 | 0.002 | 8.5237 | 6.7425 | |
| Perturbed Newtonepsilon=600.02026.07 | 0.002 | 8.5237 | 6.7425 | |
| Perturbed Newtonepsilon=400.02026.07 | 0.0025 | 6.2377 | 4.3663 | |
| Perturbed Newtonepsilon=400.02026.07 | 0.0025 | 6.2377 | 4.3663 | |
| Perturbed Newtonepsilon=200.02026.07 | 0.0036 | 3.735 | 2.1455 | |
| Perturbed Newtonepsilon=200.02026.07 | 0.0036 | 3.735 | 2.1455 | |
| Perturbed Newtonepsilon=100.02026.07 | 0.0054 | 2.3322 | 1.0504 | |
| Perturbed Newtonepsilon=100.02026.07 | 0.0054 | 2.3322 | 1.0504 | |
| Perturbed Newtonepsilon=50.02026.07 | 0.0081 | 1.4727 | 0.4827 | |
| Perturbed Newtonepsilon=50.02026.07 | 0.0081 | 1.4727 | 0.4827 | |
| Perturbed Newtonepsilon=5.02026.07 | 0.037 | 0.2902 | 0.0141 | |
| Perturbed Newtonepsilon=5.02026.07 | 0.037 | 0.2902 | 0.0141 | |
| Perturbed Newtonepsilon=0.12026.07 | 0.6197 | 0 | 0 | |
| Perturbed Newtonepsilon=0.12026.07 | 0.6197 | 0 | 0 |