AN EXPERIMENTAL COMPARISON OF THREE LEARNING MACHINE TRAINING RULES

Abstract : This report includes the results of a series of experiments to compare the efficiency of training methods using (1,0) and (+1,-1) representatives for patterns. It also presents a theoretical explanation which deals with a single TLV rather than with a network of TLV's as used in the experiments. The effects noted, however, can be expected to pertain to the network also. (Author)