Hard-margin SVM optimization on synthetic data n=10000
0.498Objective ValueGilbert Algorithm
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Gilbert Algorithmepsilon=0.001, dimension=5122017.05 | 0.498 | 2,327 | |
| Saddle-SVCepsilon=0.001, dimension=5122017.05 | 0.496 | 189 | |
| Gilbert Algorithmepsilon=0.001, dimension=322017.05 | 0.47 | 10.1 | |
| Saddle-SVCepsilon=0.001, dimension=322017.05 | 0.468 | 28.8 | |
| Gilbert Algorithmepsilon=0.001, dimension=1282017.05 | 0.438 | 152 | |
| Saddle-SVCepsilon=0.001, dimension=1282017.05 | 0.436 | 64 | |
| Gilbert Algorithmepsilon=0.001, dimension=82017.05 | 0.397 | 1.52 | |
| Saddle-SVCepsilon=0.001, dimension=82017.05 | 0.395 | 9.33 |