Machine Unlearning on Credit
0.046Unlearning Runtime (s)EUPG (ϵ=1, δ=0)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| EUPG (ϵ=1, δ=0)Model Family=XGBoost, Forget ratio=5%, Number of runs=52025.07 | 0.046 | 2.46 | |
| EUPG (k=30)Model Family=XGBoost, Forget ratio=5%, Number of runs=52025.07 | 0.048 | 2.74 | |
| RetrainModel Family=XGBoost, Forget ratio=5%, Number of runs=52025.07 | 0.113 | 1 | |
| SISAModel Family=XGBoost, Forget ratio=5%, Number of runs=52025.07 | 0.661 | 0.33 | |
| EUPG (ϵ=1, δ=0)Model Family=Neural Network, Forget ratio=5%, Number of runs=52025.07 | 5.55 | 14.84 | |
| Certified-SP (ϵ=1, δ=1e-5)Model Family=Neural Network, Forget ratio=5%, Number of runs=52025.07 | 5.59 | 14.04 | |
| EUPG (k=30)Model Family=Neural Network, Forget ratio=5%, Number of runs=52025.07 | 5.59 | 14.76 | |
| RetrainModel Family=Neural Network, Forget ratio=5%, Number of runs=52025.07 | 52.54 | 1 | |
| SISAModel Family=Neural Network, Forget ratio=5%, Number of runs=52025.07 | 82.88 | 0.62 |