Difference Variation Of Clonal Selection Algorithm And Its Application

Biological immune system is an adaptive system, which can learn, memory, identify and extract feature. The system can effectively recognize and eliminate invading antigens. Artificial immune algorithm (AIS) is a novel computational framework inspired by biological immune systems. Clonal Selection Algorithm (CSA) is one of important branches of AIS, and it is used to solving difficult high-dimensional optimization problems. Clonal selection and hypermutation mechanisms are its main characteristics.According to the features of higher-dimensional Rotated and shifted functions, this paper presents a novel heuristic algorithm, name Greedy Immune Memory CSA, to prevent the evolutionary stagnation of the population. The immune memory mechanism is employed to strength the exploitation ability of population. Meanwhile, a multi-mutation strategy and a multi-round competition are adopted. A serial experiment carries out through the testing suit of CEC2005, and the results show the presented algorithm improves the solution quality.When dealing with global optimization problems, immune algorithm faces the problem of insufficient diversity. This paper incorporates differential evolution into the operation of clone mutation, and proposes a new improved clonal selection algorithm,called DECSA (clonal selection algorithm based on differential evolution), which combines differential evolution with clonal super-mutation. This method promotes the exchange of information between antibody and antibody, lets offspring inherit their parent antibody’s information and carry other parent antibody’s information at the same time and, as a result, enriches the diversity of antibody populations. This method can perform global search and local search in many directions rather than one direction around the identical antibody simultaneously. The results show that the algorithm can maintain the population diversity algorithm, especially in the pre-search algorithm has very fast convergence rate, to avoid premature convergence algorithm also has some contribution, to ensure convergence of the algorithm, but also effectively improve the accuracy of the search solution to enhance the robustness.Clonal selection algorithm using differential variation of the RBF neural network was trained and optimized using a modified differential evolution of neural networks to approximate the nonlinear system. The improved algorithm of the RBF neural network center value, width, and weight were optimized. Simulation results show that the differential mutation clonal selection algorithm more than the genetic algorithm approximation ability of nonlinear systems.