Immune Algorithm with Adaptive Sampling in Noisy Environments and Its Application to Stochastic Optimization Problems

An immune optimization algorithm in noisy environments, suitable for high-dimensional stochastic optimization problems, is proposed based on the hypothesis test and simplified immune metaphors of c. The focus of design is concentrated on constructing three types of operators: (1) population sampling that decides sampling sizes of both the current population and the memory set, (2) sample-allocation scheme, and (3) antibody evolution that is aimed at designing several immune operators to evolve some potential antibodies into better ones. The algorithm, depending on dynamic suppression radiuses and suppression probabilities of antibodies from evolving populations, can strongly suppress noise and rapidly discover the desired solution, even if prior information on noise is unknown. Experimental results and comparison with three well-known algorithms show that the proposed algorithm can achieve satisfactory performances including the quality of optimization, noise compensation and performance efficiency.