Algorithm of Recurrent Multilayer Perceptrons Learning for Global Modeling of Complex Systems

In this paper, the problem of modeling and identification of complex input-output systems using recurrent neural networks is discussed. In such a system, we can distinguish a sub process (elementary process) with some inputs and outputs, which can operate separately. Connection between inputs and outputs of each element gives us a complex system. Each element of the complex system is modeled by one multi-input multi-output neural network. We obtain model of the whole system by composing all neural networks into one global network. Existing learning algorithms of recurrent neural networks do not provide solution of this problem. Main contribution of this work is an algorithm based on gradient descent method for identification of dynamic properties of a complex system with cascade structure using Recurrent Multilayer Perceptrons.