The amount of data available for training a neural net is often limited, and obtaining additional data can be expensive. The new Testing Sensitivity Analysis in version 6 helps to make the most of small data sets.
When a neural net is trained on a small data set, the subset of data that is used to test the neural net is also small, which limits the reliability of the testing results. This new analysis helps to determine if testing results are reliable given the amount of data that has been set aside for testing. It can also answer the question whether changing the size of the testing subset will increase the reliability of the results.