Table 1.
Comparison of different algorithms
| Method | Single breakpoint | 1000-1000 | 500-1500 | 300-1500 |
|---|---|---|---|---|
| Viterbi | 7 (3..17) | 6 (2..17) | 7 (2..20) | 7 (3..16) |
| max ball, r = 1 | 6 (2..16) | 6 (2..18) | 7 (1..19) | 7 (2..16) |
| max ball, r = 2 | 5 (2..16) | 6 (1..17) | 6 (2..18) | 6 (2..15) |
| max ball, r = 5 | 4 (1..12) | 5 (3..14) | 5 (3..15) | 5 (2..13) |
| max ball, r = 7 | 4 (2..12) | 5 (3..14) | 6 (3..14) | 6 (3..12) |
| max ball, r = 10 | 6 (3..10) | 7 (3..14) | 8 (3..14) | 7 (3..10) |
| max ball, r = 15 | 7 (3..11) | 9 (4..14) | 10 (5..15) | 7 (4..12) |
| min region | 4 (1..13) | 5 (2..14) | 6 (2..16) | 5 (2..14) |
| DualBrothers | 57(45..71) | 51 (42..67) | 48(37..54) | 45 (40..51) |
Shown are the median and interquartile range of differences between predicted and actual breakpoints for the four initial experiments.