View the overall training, validation, and test loss.
Select the Loss Metrics radio button.
This loss is a constant value that represents the error made by the trained romAI model when predicting over the testing dataset. The data is
reported in the text box of the loss chart and refers to the value calculated after the
training as shown below:
The data also shows the training/validation loss of each single output or higher-order
state variable, such as velocity in this case.
At this point, the typical machine learning considerations are valid, such as:
The lower the training error, the more accurate the ROM.
If the training and test loss is close to each other, then the training has
converged.
If the training and test losses diverge, then the ROM might be affected by
overfitting. In this case, you may want to consider increasing the
regularization term.
The cross-validation ratio is used to monitor the Early Stop criteria. To change
the cross-validation ratio, select the Early Stopping check box as shown
below.
