*6.2. Neural Network*

The neural network used is an NLARX with two hidden layers of 25 neurons each. The output for the validation can be found in Figures 9 and 10. Like the SM case, the resulting fit value obtained exceeds expectations and some concerns of over-fitting arise. The fitted values for pith and roll are 99.0162% and 99.2027%, respectively (see Table 3). Similarly to the behavior of the SM case, the training data for the neural network covers the validation data with high precision. Therefore, additional tests are required to properly evaluate the model performance.
