Table 4:
Summary of score contributions for 2-way epistasis dataset in Table 2 yielding ReliefF, SURF, SURF*, MultiSURF*, and MultiSURF scores. Illustrates why ‘far’ scoring improves 2-way interaction detection.
Features | RBA Scoring Schemes | |||||||
---|---|---|---|---|---|---|---|---|
A1 | A2 | A3 | RelielF | SURF | SURF* | MultiSURF* | MultiSURF | |
Farthest Hit (same value) | 0 | 0 | 0 | − 1 | ||||
Farthest Miss (same value) | 4 | 4 | 0 | +1 | ||||
Nearest Hit (different value) | 0 | 0 | 8 | − 1 | − 1 | − 1 | − 1 | − 1 |
Nearest Miss (different value) | 4 | 4 | 0 | +1 | +1 | +1 | +1 | +1 |
Farthest Hit (different value) | 8 | 8 | 8 | +1 | ||||
Farthest Miss (different value) | 4 | 4 | 8 | − 1 | ||||
RelielF Score Total | 4 | 4 | −8 | |||||
SURF Score Total | 4 | 4 | −8 | |||||
SURF* Score Total | 8 | 8 | −8 | |||||
MultiSURF* Score Total | 8 | 8 | −8 | |||||
MultiSURF Score Total | 4 | 4 | −8 |