Nonlinear system identification based on evolutionary dynamic neural networks with complex weights

The paper presents a novel dynamic neural architecture that allows a flexible and compact representation of nonlinear processes. The suggested neural topology is obtained by providing local internal recurrence for the static neural network with complex weights. An evolutionary multiobjective design procedure assists the automatic selection of appropriate neural topologies and parameters. It searches for accurate neural models, characterised by good generalisation capabilities. The experiments reveal that the presented approach is suitable for system identification.