The example neural network included with NNCS simulates the first two levels of visual information processing that the eye does. The first level is the retina and the second level is a layer of ganglion cells that functiona as a contrast detector. The retina is the input level, or the level in which the data originates. I use a set of 10 neurons labeled "Retina 1," "Retina 2," etc. to represent the retina. The resting values of these cells reflect the intensity of light arriving at a corresponding position of the eye. It is set up now with one unit of activation in Retinal Cells 1-5 and five units of activation in Retinal Cells 6-10. Thus, the current pattern is a simple edge. You can examine the effects of different patterns by changing the resting levels of cells in the retina.
The second-level cells, 8 neurons labeled "Ganglion 1," "Ganglion 2," etc., are specialized to detect changes in the retina. Each ganglion cell is excited by a single cell (called its center) in the retina and inhibited by the the two surrounding cells (called its surround). Thus, a ganglion cell reacts to differences in intensity between its center and its surround. If the center and the surround are equally active, the ganglion cell will exhibit no activity because the excitation and inhibition will cancel each other out.
Clear the network and run it for one cycle to see how the ganglion cells react to the edge in the retina. Ganglion Cells 1-3 and 6-8 have no activity since there is no contrast between their centers and surrounds. Ganglion 4, whose center is just before the edge, has an activity level of -4, and Ganglion 5, whose center is just after the edge, has an intensity level of +4. This characteristic pattern is called a "zero-crossing" because activity in the ganglion-cell layer dips and crosses zero before returning to zero.
So I hope this little network will give you something to play around with. It only has a retina of ten cells, so it certainly can be expanded. Also, below is a bibliography which will give you more ideas, and better in-depth explanantions (A reprint of the one written by me can be obtained directly if you write me a request.).
Halff, L. A. (1987). A Spreadsheet Simulation of the Early Stages of
Visual Information Processing. Behavioral Research Methods,
Instruments, and Computers, 19, 117-122.
Lindsay, P., & Norman, D.A. (1972). Human Information Processing. New
