Combine several models by weighted voting or weighted averaging to get an ensemble probability, and see how each model weight shifts the result.
权重可基于模型验证性能分配。
| Item | Formula / Note | Source / Standard |
|---|---|---|
| Weighted fusion | fused probability sums model outputs weighted and normalised | Ensemble learning |
| Soft and hard voting | soft voting averages probabilities while hard voting counts labels | Ensemble learning |
| Diversity gain | gains grow when member errors are less correlated | Ensemble learning |