A Concis Presentation of Supiervised Learing Algorithems for Feedforward Neural Netwoks

Abstract

With an increasing interest in neural networks for identification and control of dynamic systems, there is a need to present neural network training algorithms in a form suitable for immediate implementation into a computer code by new students of neural networks, whether in academe or in industry. The purpose of mis paper is to fulfill this need. A concise presentation of standard back propagation and extended Kaiman filter based algorithms is given. The work presented here was carried out at Ford Motor Company during a sabbatical leave.