A Cluster and Gradient-Based Artificial Immune System Applied in Optimization Scenarios

The main objective of this paper is to use artificial immune systems (AIS) in optimization problems. For this purpose, two major immunological principles presented in the literature are revisited: hypermutation, which is responsible for local search, and receptor edition, used to explore different areas in the solution space. This paper presents three major modifications divided into two different goals. The first goal is to speed up the convergence of each individual. This is done through a new hypermutation approach that uses the numerical information provided by the optimization system to drive the cloning process to interesting directions into the solution space. The second goal regards the reduction of the computational effort necessary to simulate the whole population. This is accomplished by adding to the AIS algorithm two more features of the natural immune system: maturation control and memory cells. The maturation control analyzes the antibodies and, during the convergence process, eliminates possible redundancies, represented by individuals driving to the same local optimum. The last proposed improvement is the use of memory cells in dynamic-optimization scenarios. In such situations, a repertoire of successful cases is used to forecast part of the initial population. Combining these concepts together decreases the number of antibodies, generations, and clones, consequently speeding up the convergence process. Applications illustrate the performance of the proposed method.