Table 2.
Ratio | Number | MH-L1SVM | MH-L2SVM | L1SVM | L2SVM |
---|---|---|---|---|---|
1 | 600, 404 | 29 ± 1 | 28 ± 1 | 188 ± 32 | 387 ± 63 |
5 | 1, 801, 212 | 172 ± 5 | 38 ± 2 | 1, 655 ± 156 | 963 ± 81 |
10 | 3, 302, 222 | 448 ± 41 | 261 ± 7 | 1, 261 ± 579 | 10, 798 ± 1, 981 |
25 | 7, 805, 252 | 1, 808 ± 181 | 732 ± 17 | 20,067 ± 1,453 | 4, 623 ± 782 |
50 | 15, 310, 302 | 1,140 ± 90 | 811 ± 41 | 58, 045 ± 5, 678 | 8, 936 ± 1, 412 |
100 | 30, 320,402 | 7, 601 ± 627 | 1,643 ± 50 | > 24hours | 16, 608 ± 2, 732 |
250 | 75, 350, 702 | 25,060 ± 12,417 | 4,631 ± 795 | > 24hours | 43, 843 ± 7, 200 |
The training time includes minwise hashing time. The number of negative examples is varied from the same number of positive examples to the number of negative examples 250 times larger than the number of all positive examples.