TABLE II.
Unsupervised classification accuracy for parameter values using k-medoid clustering with various input feature vectors and k = 25. The third column displays the accuracy when simulations are classified by parameter vector (C, ℓ), and the fourth column displays the accuracy when simulations are classified by phenotype.
| Summary | Feature | Parameter (%) | Phenotype (%) |
|---|---|---|---|
| Order parameters | P(t) | 20.6 | 62.9 |
| Mang(t) | 19.3 | 60.7 | |
| Mabs(t) | 47.3 | 63.9 | |
| DNN(t) | 48 | 80.1 | |
| All | 49.9 | 97.4 | |
| TDA (position) | b0 | 76.6 | 83.6 |
| b1 | 71.9 | 79.8 | |
| b0 & b1 | 76.6 | 99.8 | |
| TDA (time-delayed position) | b0 | 71.1 | 82.6 |
| b1 | 73.7 | 84.4 | |
| b0 & b1 | 71.3 | 99.0 |