Research Article
Comparison of Two Quantitative Analysis Techniques to Predict the Evaluation of Product Form Design
Table 4
RMSE of NN prediction models for automobile profile training dataset.
|
Number of training epochs | NN-Arith. | NN-Geom. | NN-Sum | I: 92 neurons H: 49 neurons O: 6 neurons | I: 92 neurons H: 23 neurons O: 6 neurons | I: 92 neurons H: 98 neurons O: 6 neurons |
| 1000 | 0.1028 | 0.1031 | 0.1069 | 2000 | 0.0956 | 0.0970 | 0.0964 | 5000 | 0.0846 | 0.0871 | 0.0872 | 10,000 | 0.0775 | 0.0787 | 0.0791 | 20,000 | 0.0672 | 0.0685 | 0.0684 | 30,000 | 0.0620 | 0.0638 | 0.0643 | 40,000 | 0.0588 | 0.0606 | 0.0610 | 50,000 | 0.0576 | 0.0599 | 0.0603 | 60,000 | 0.0567 | 0.0587 | 0.0592 | 70,000 | 0.0558 | 0.0577 | 0.0584 | 80,000 | 0.0553 | 0.0570 | 0.0579 | 90,000 | 0.0550 | 0.0570 | 0.0577 |
|
|
“I,” “H,” and “O” indicate the input, hidden, and output layer, respectively.
|