Secondary Immune Response for Evolutionary Time Dependent Optimization

In this paper, we investigate the relevance of two simple computational models of immunization in Time Dependent optimization (TDO) problems. At rst, we propose a Simple Arti cial Immune System (Sais) and experimentally compare its reactiveness and robustness to well known evolutionist approaches. Sais is then applied to a cyclic dynamical environment in order to evaluate its ability to feature an improved robustness when facing previously encountered optima. After discussing the limits of this approach, we propose a second algorithm (Yasais) designed to improve this so-called immunization process by stabilizing the way optima are memorized. Eventually, we discuss the results of both algorithms and underline how the latter features a quasi-optimal behavior.