This panel specifies how to train the network and when to stop the simulations. There are separate panels for supervised and unsupervised learning. The figure below illustrates the supervised case.
The stop criterion can be based on either the number of training epochs or the mean square error (MSE), whichever happens first. The MSE stop criterion can use either the training set or test set. When cross validation is specified from the Network Analysis panel, the default configuration will stop the network when the MSE of the test set begins to rise (meaning that the network has just begun to overtrain).
This panel also specifies when the weights are updated. On-line learning updates the weights after the presentation of each exemplar. In contrast, batch learning updates the weights after the presentation of the entire training set.
NeuroSolutions Home Page