Table 1 The table shows the results of dropping data that falls in the margin. For any data point xi; let maxm{fm(xi)} = fyi, and Let fm = maxm{fm(xi)} for all m ≠ yi, then we define the margin as: (fyi - fm), hence data point xi is dropped if (fyi - fm) ≤ Confidence Parameter. Using the margin drop approach, there is even less tuning, and there is improved throughput (approximately 75% for all species).
From: Support Vector Machine Implementations for Classification & Clustering
Kernel | 8GC | 9AT | 9CG | 9GC | 9TA |
|---|---|---|---|---|---|
Gaussian | P: 1268 TP: 1087 SN+SP: 1.76 P: 1087 TP: 1087 SN+SP: 2 Drop = 9.42 | P: 1178 TP: 934 SN+SP: 1.57 P: 934 TP: 934 SN+SP: 2 Drop = 22.17 | P: 1166 TP: 904 SN+SP: 1.53 P: 904 TP: 904 SN+SP: 2 Drop = 24.67 | P: 1172 TP: 897 SN+SP: 1.51 P: 897 TP: 897 SN+SP: 2 Drop = 25.25 | P: 1216 TP: 1027 SN+SP: 1.70 P: 1027 TP: 1027 SN+SP: 2 Drop = 14.42 |
AbsDiff | P: 1407 TP: 1134 SN+SP: 1.75 P: 1134 TP: 1134 SN+SP: 2 Drop = 5.5 | P: 1151 TP: 928 SN+SP: 1. 58 P: 928 TP: 928 SN+SP: 2 Drop = 22.67 | P: 1177 TP: 906 SN+SP: 1.53 P: 906 TP: 906 SN+SP: 2 Drop = 24.5 | P: 1050 TP: 870 SN+SP: 1.55 P: 870 TP: 870 SN+SP: 2 Drop = 27.5 | P: 1215 TP: 1040 SN+SP: 1.72 P: 1040 TP: 1040 SN+SP: 2 Drop = 13.33 |
Entropic | P: 1165 TP: 1038 SN+SP: 1.75 P: 1038 TP: 1038 SN+SP: 2 Drop = 13.5 | P: 1480 TP: 995 SN+SP: 1.50 P: 991 TP: 991 SN+SP: 2 Drop = 17.42 | P: 1348 TP: 922 SN+SP: 1.45 P: 920 TP: 920 SN+SP: 2 Drop = 23.33 | P: 960 TP: 804 SN+SP: 1.50 P: 803 TP: 803 SN+SP: 2 Drop = 33.08 | P: 1047 TP: 970 SN+SP: 1.73 P: 970 TP: 970 SN+SP: 2 Drop = 19.17 |