A Novel Method of Neural Network Optimized Design Based on Biologic Mechanism

Optimized design of neural network based on biologic immune modulated symbiotic evolution (BIM) is proposed, which combines with the adjustment of antibody of immune modulated theory so as to keep the individual diversity. With combining evolved intergrowth algorithm and density of immune principle suppress modulation mechanism together, system shortens the individual's length of code and lightened the calculating amount by solving the evolution of the colony to the neuron part, eliminates the premature convergence effectively. Meanwhile, system adopts the improved immune adjustment algorithm, which improved the variety of the colony availably. The neuron that produced in the colony in this way can get and realize the network quickly. The results of simulation experiment which applies in system of the two stands reversing tandem cold mill show that this method is applied to the complicated climate, it has good capabilities of convergence and capability of resisting disturbance.