Automatic Generation of Neural Network Structures for Modeling the Highly Nonlinear Processes

M. Spišiak and Š. Kozák (Slovakia)


Artificial neural network, genetic algorithm, occurrence matrix


This paper deals with a generalized automatic method used for designing artificial neural network (ANN) structures. ANN's are applied in modeling and control of highly nonlinear large-scale processes. One of the most important problems is to design the optimal ANN for many real applications. In many applications, optimal ANN structures are designed by heuristic approaches or simply determined by experiments. In this paper, two techniques for automatic finding an optimal ANN structure are proposed. They can be applied in real-time applications as well as in fast nonlinear processes. One possible approach proposed in this paper consists in using the genetic algorithms (GA). It can be used to find an optimal or a minimal ANN structure. The first proposed method deals with designing a structure with one hidden layer. The optimal structure has been verified on a nonlinear model of a hydraulic system. The second algorithm allows to designs ANN with an unlimited number of hidden layers each of them containing of one neuron. This structure has been verified on a highly nonlinear model of a polymerization reactor. The obtained results have been compared with the results yielded by fully connected ANN.

Important Links:

Go Back