Table 8. Pareto set of different methods on the USPS dataset.
| Method | Reduced (ratio) | Accuracy | Training Speedup | Testing Speedup | Reduced SVs (ratio) | Rank |
|---|---|---|---|---|---|---|
| BRIX (r = 0.02) | 0.02 | 0.896 | 72.90 | 6.77 | 0.10 | 1 |
| BRIX (r = 0.04) | 0.04 | 0.915 | 62.17 | 5.78 | 0.14 | 2 |
| BRIX (r = 0.03) | 0.03 | 0.909 | 63.79 | 5.85 | 0.13 | 3 |
| BRIX (r = 0.05) | 0.05 | 0.917 | 51.77 | 5.12 | 0.16 | 4 |
| BRIX (r = 0.1) | 0.10 | 0.920 | 33.35 | 4.84 | 0.18 | 5 |
| BRIX (r = 0.01) | 0.01 | 0.786 | 98.19 | 9.44 | 0.07 | 6 |
| BRIX (r = 0.2) | 0.20 | 0.921 | 18.85 | 4.38 | 0.20 | 7 |
| SVOX (k = 2) | 0.10 | 0.918 | 34.48 | 4.18 | 0.22 | 8 |
| BRIX (r = 0.3) | 0.30 | 0.921 | 12.87 | 4.15 | 0.23 | 9 |
| BRIX (r = 0.5) | 0.50 | 0.921 | 6.57 | 3.00 | 0.31 | 10 |
| BRIX (r = 0.6) | 0.60 | 0.922 | 5.05 | 2.77 | 0.34 | 11 |
| BRI (r = 0.5) | 0.50 | 0.922 | 5.14 | 2.17 | 0.46 | 12 |
| BPLSH (M = 50;L = 20) | 0.63 | 0.922 | 1.64 | 1.14 | 0.92 | 13 |
| CBCH (k = 700) | 0.67 | 0.922 | 1.54 | 1.05 | 0.96 | 14 |
| SE (r = 0.7) | 0.70 | 0.922 | 1.46 | 1.04 | 0.97 | 15 |
| SE (r = 0.8) | 0.80 | 0.922 | 1.30 | 1.04 | 0.98 | 16 |
| BPLSH (M = 90;L = 30) | 0.83 | 0.922 | 1.20 | 1.04 | 0.96 | 17 |
Methods in the Pareto set are ordered by their rank according to closeness to the optimal solution. “Reduced (ratio)” is the ratio of the size of the resulting reduced dataset to the size of the whole dataset. “Reduced SVs (ratio)” is the ratio of the number of SVs of the reduced dataset to that of the whole dataset.