A NEW SYSTEM-IDENTIFICATION TECHNIQUE USING KALMAN FILTERING AND MULTILAYER NEURAL NETWORKS

The objective of this work is to use the back-propagation algorithm in conjunction with Kalman filtering in order to establish a new self-le arning technique of multilayer neural network (MNN). This new techniqu e is developed by directly building a Kalman filtering model for each perceptron in order to increase the adaptability of the MNN and to pro vide for on-line nonlinear system identification. We demonstrate that this new technique is faster and more stable than the classical back-p ropagation algorithm for training multilayer perceptrons. We also find that it is less sensitive to the initial weights and to the learning parameters.