Once you have defined the inputs and outputs for a problem, you need to define the neural network’s Architecture and Training Criteria. Double-clicking in this cell presents you with the familiar Design module from NeuroShell 2. From here you may select any of the available neural network paradigms and architectures and set the training criteria. This gives you the power to modify a single base configuration for multiple aliases of a problem.
If a configuration file (.FIG) exists for the problem name, it will appear in the Design module. Exiting the Design module will place the existing .FIG file for your problem into this cell. This is to show that a configuration has been selected for the problem.
For example, you may want to test whether a three-layer Backpropagation network is better than a recurrent network. You would reference the configuration file from the base problem and modify the network type to a recurrent network for use in the second (alias) problem.