Advanced exercise: Modify the program to model tasks in which a Kohonen network deals with a highly irregular probability of points coming from different regions of an input signal space. Conduct self-learning and you will see that in a taught network, many more neurons will specialize in recognizing signals coming from regions more often represented in a teaching set. Compare the effect with a map of a cerebral cortex (Figure 10.40). What conclusions can you draw from this exercise?
