A selective learning algorithm for certain types of learning failure in multi-layer perceptrons

Summary form only given, as follows. A simple selective learning algorithm for use with multilayer perceptrons is presented. This algorithm has proved useful in certain types of problems where learning failure occurs using standard backpropagation. Examples of these problems are included. The algorithm is based on the RMS output error computed across all output nodes and all training patterns. The learning rate is decreased for all individual output nodes each time the error is less than a user-chosen multiple of the RMS error corresponding to the previous pass. This algorithm has produced convergence where the standard fixed-again backpropagation failed.<>