In some cases, when to save the trained network depends upon whether or not you're using Calibration, which is defined as having the Calibration interval set greater than 0 and having a test set. See Backpropagation -Calibration for details.
Click on one of the following buttons to determine when to save the network:
Best Training Set - saves the network every time it reaches a new minimum average error for the training set. You may want to use this option if you're NOT using Calibration. (When using this option, the minimum average error computations are done at the end of each epoch.)
Best Test Set - saves the network every time it reaches a new minimum average error for the test set. Select this option if you're using Calibration. (The computations for the test set are done at end of the specified number of events.)
When using Calibration, the network is applied to the test set at the number of events specified for the test interval, and the mean squared error is computed for the data in the test set. Whenever the mean squared error is less than any previous mean squared error, the network is automatically saved if you have selected Best Test Set.
No Auto Save - does not save the network. You may want to use this option to speed up training at the beginning when the network is constantly finding new minimum errors, but you need to save the network before closing the module. When new minimum errors are found less frequently, you may want to turn off No Auto Save and select either Best Training Set or Best Test Set.
You may change any of the save options during training.