Table 3
Comparison of ANNs performance with 11 different BPLAs to predict C DCM.
| BPLA | Function | MSE × 103 | R 2 | Epochs |
|---|---|---|---|---|
| Resilient backpropagation | Trainrp | 0.47245 | 0.99672 | 2000 |
| BFGS quasi-Newton backpropagation | Trainbfg | 0.11915 | 0.999 | 1900 |
| Conjugate gradient backpropagation with Powell-Beale restarts | Traincgb | 0.19062 | 0.99847 | 777 |
| Conjugate gradient backpropagation with Fletcher-Reeves updates | Traincgf | 0.44301 | 0.99723 | 308 |
| Conjugate gradient backpropagation with Polak-Ribiere updates | Traincgp | 0.16445 | 0.9988 | 1275 |
| Gradient descent with adaptive learning rate backpropagation | Traingda | 1.339 | 0.99103 | 4992 |
| Gradient descent with momentum backpropagation | Traingdm | 9.0481 | 0.939 | 5000 |
| Gradient descent with momentum and adaptive learning rate backpropagation | Traingdx | 0.78019 | 0.99426 | 5000 |
| Levenberg-Marquardt backpropagation | Trainlm | 0.05714 6 | 0.99954 | 479 |
| One-step secant backpropagation | Trainoss | 0.17262 | 0.99866 | 4008 |
| Scaled conjugate gradient backpropagation | Trainscg | 0.11095 | 0.99921 | 4870 |