Reading and Writing Files

SNNS supports five types of files, the most important ones are:

Network definition files containing information on network topology and learning rules. The files end in the extension '.net'.

Pattern files, containing the training and test data. All pattern files end in '.pat'.

Results files. Network output is interpreted in many possible ways, depending on the problem. SNNS allows the user to dump the network outputs into a separate file for later analysis.

The first thing you are likely to use is the option in the manager panel to read network and pattern definition files. The window that will appear is given in figure .

  
Figure: The SNNS file browser

The top text field shows the current directory. The main field shows all files for each of the file types that SNNS can read/write. Directories are marked by square brackets. To load an example network change the directory by entering the example directory path in the top field (do not press return):

SNNSv4.1/examples

Changes will only be apparent after one of the file selectors has been touched:click on and then again. You should now see a list of all network definition files currently available. Double-clicking on one of the filenames, say 'letters' will copy the network name into the file name window. To load the network simply click on . You can also enter the filename directly into the file name window (top left).