The Pi-sigma network learning algorithm is generalized
and a simple method for computing product of fuzzy numbers is proposed
so that it is possible to utilize both numerical patterns and expert knowledge represented by fuzzy IF-THEN rules for designing the Pi-sigma network classifiers. Simulation results indicate that the generalized learnig algorithm is correct
and in some cases
the method based on the Pi-sigma network requires simpler structure and much shorter training time than that based on the Multilayer Perceptron.