Abstract:

The actually complicated objects are difficult to control due to their unknown mathematical models. In order to overcome this difficulty, a pattern recognition algorithm based on the artificial immune network is proposed, in which the discrete models and learning algorithms of artificial immune network are adopted and the functions of artificial immune system are combined with the framework of artificial neural network. After that, an objectmodel based on artificial immune network is constructed. The proposed algorithm merges the positioning and para- meter adjustment of network nodes as well as the parameter tuning of basis functions, etc. Thus, the two stage tasks of Radial Basis Function （RBF） neural network are effectively accomplished and the recognition accuracy is greatly improved. Simulated results in terms of two object functions finally confirm the high convergence speed and great accuracy of the proposed algorithm.

Key words: artificial immune network, pattern recognition, network structure, parameter optimization, node-parameter adjustment