Abstract
A generalization of the potential function method is made for multi-class pattern classification. Based on the construction of a set of separation functions an algorithm is proposed for teaching automata to classify various classes of input patterns. Using a basic assumption that the potential functions are bounded and the separation functions can be represented by a linear combination of a set of certain real scalar functions, the convergence of the algorithm is proved.
Notes
† Communicated by Professor Wah-Chun Chan. This work was supported by the National Research Council of Canada under the Grant A-5127.