Table 3 Summary of the training results for the five best ANNs in the RCMSC-ANN variant of the analysis. See text for details.
ID | ANN | Prediction quality [%] | Training algorithm and number of epochs | Training error function | Activation function of the hidden neurons | Activation function of the output neurons | ||
|---|---|---|---|---|---|---|---|---|
T | U | V | ||||||
1 | MLP 7–12-2 | 91.43 | 91.43 | 92.86 | BFGS 4 | ESOS | Exponential | Sigmoid |
2 | MLP 7-7-2 | 97.14 | 95.71 | 98.57 | BFGS 11 | ESOS | Sigmoid | Identity |
3 | MLP 7-4-2 | 91.43 | 94.29 | 95.61 | BFGS 7 | ESOS | Sine | Tanh |
4 | MLP 7–9-2 | 97.14 | 92.86 | 98.57 | BFGS 7 | ESOS | Tanh | Tanh |
5 | MLP 7–9-2 | 97.14 | 94.29 | 98.57 | BFGS 10 | ESOS | Sigmoid | Identity |
