Research On Immune Genetic Neural Networks

In the last few years we could perceive a great increase in studying biologically inspired system. Among these, we can emphasize neural networks, evolutionary computation, DNA computation, and immune system. Immune genetic neural networks is a complex of immune algorithem, genetic algorithm and neural networks which has proven to be capable of performing several tasks, lide pattern recognition, learning, memory acquisition, generation of diversity, noise tolerance, generalization, distributer detection and optimization. Based on biological principles, new computation techniques are being developed, aiming not only at a better understanding of the sysytem, but also at solving engineering problem.Referring to the concept and theory immune in biotic science, we do research on the theory, algorithm and application of Immune genetic neural networks. First, a simple introduction of immune algorithem is offered, and the comparison of characters with ANN and GA is made. Second, an optimizing method based on immune genetic algorithm for designing multilayer feed-forward neural network is presented in this paper. This algorithm can decide the structure of the multilayer feed-forward neural networks and search the proper weight of the networks. The simulation experiments show that this algorithm has the better ability of convergent on whole solution space and the capacity of fast learning. Third, according to these merits, we bring them into a state of unity. Absorbing the merits of multiple-valued immune network from Zhang Tang and ABNET from De Castro, a new computation model—Immune genetic neural networks is presented. The core of this model is that it refers to the principle of immunology and uses the structure and theory of ANN, defines the basic computation units and rules based on immunology and uses GA to adjust the dynamic balance of generation's convergence and individual's diversity. This model simulates a lot of theory of immunology on engineering. Compared with other methods, its application to solve the simple clustering problem and corks evaluating problem shows that this model has the better ability of cluster and capacity of convergent and fast learning.