AN OPTIMIZATION METHOD FOR THE TOPOLOGICAL STRUCTURES OF FEED-FORWARD MULTI-LAYER NEURAL NETWORKS

摘要

An empirical formula for initially selecting the number of hidden nodes of supervised linear basis function (LBF) feedforward three-layered neural networks is summarized and introduced according to many applications in various fields, and a new method for judging whether or not there exist redundant nodes in the hidden layer of such a neural network proposed. Finally, some examples are given to verify them.