Building Model

This is the button that has to be pressed to both begin and end the learning process.

The algorithm determines in which direction the model 'weights' need to change in order to decrease the error, the learning rate determines the magnitude of the change. Adjust the learning rate while the model is being trained to see the effect of this setting on the plot.

This displays the current error and the best error to date. The error can actually increase if the learning rate is set too high.

This setting determines how often the graphs are refreshed.

Replotting the graph takes up time. If you are not interested in the visuals then deselect this check box.

This will reset the model weights to a random starting point. This may need to be done if the learning appears to have stalled.

This will increase the number of hidden neurons in the neural network. More hidden neurons gives the model the ability to learn more complex relationships but also the ability to learn spurious noise.

The data can be partitioned into a training and validation set. Pressing this button allows the training data to be specified.

When the model training has been stopped, there is the option to select the model that performed best on the training data or best on the validation data. This is accessed from the neural drop down menu.