Computational Intelligence in Electromagnetics

A Combined use of the three techniques globally referred to as computational intelligence (i.e., Artificial Neural Networks, Fuzzy Logic and Evolutionary Computation) is proposed to cope with the solution of inverse problems. Electromagnetics is presented as an interesting application field of the methods. Firstly, it is shown how simple autoassociative ANN might successful play a role in fault tolerant feature extraction. Secondly, hybrid neuro-fuzzy systems are shown to have better reconstruction performance with respect to ANN and FL systems working alone. Finally, evolutionary techniques are used to improve the strategy of optimization of a neural processor in a simple identification problem.