Artificial Immune Systems: Principle, Algorithms and Applications

Abstract

The present thesis aims to make an in-depth study of adaptive identification, digital channel equalization, functional link artificial neural network (FLANN) and Artificial Immune Systems (AIS).Two learning algorithms CPSO and IPSO are also developed in this thesis. These new algorithms are employed to train the weights of a low complexity FLANN structure by way of minimizing the squared error cost function of the hybrid model. These new models are applied for adaptive identification of complex nonlinear dynamic plants and equalization of nonlinear digital channel. Investigation has been made for identification of complex Hammerstein models.
To validate the performance of these new models simulation study is carried out using benchmark complex plants and nonlinear channels. The results of simulation are compared with those obtained with FLANN-GA, FLANN-PSO and MLP-BP based hybrid approaches. Improved identification and equalization performance of the proposed method have been observed in all cases.