Input Values Function for Improving Generalization Capability of BP Neural Network

Abstract

As is known to all, Back propagation (BP) neural network has two important advantage and disadvantage: learning speed and generalization capability (GC). In this paper, we propose a new method by adding input values function (IVF) to improve the generalization of BP neural network. The result shows: GC in a certain extent has been improved throught this method. But if we want to do much more promoting in various fields, there are still a lot of things that we must to be studied.