Two- and Three-Layer Recurrent Elman Neural Networks as Models of Dynamic Processes
The goal of paper is to study and compare the effectiveness of two- and three-layer Elman recurrent neural networks used for modelling of dynamic processes. Training of such networks is discussed. For a neutralisation reactor benchmark system it is shown that the rudimentary Elman structure with two layers is much better in terms of accuracy and the number of parameters. Furthermore, its training is much easier.