This problem consists of classifying a set of randomly generated patterns (with real values) in two classes. An example in two dimensions is given by the ``healthy food'' learning problem. Inputs are given by points in the ``smell'' and ``taste'' plane, corresponding to the different foods. The learning task consists of producing the correct classification as ``healthy food'' or ``junk food'' (Figure 9.25).
On this problem we obtained a speedup of 20-120 (going from 6 to 100 patterns in two dimensions).