Why Are Clonal Selection Algorithms MCMC?
K. El-Nady, Usama Abou El-Enien, Amr Ahmed Badr · Zenodo (CERN European Organization for Nuclear Research) · 2011
AbstractClonal selection algorithms are considered. Two algorithmsare designed and executed to obtain purely empirical analy-sis conclusions in order to turn to purely theoretical analy-sis results about the behavior of clonal selection algorithmsas Markov chains, which con¯rm the conjectures from theseexperiments and in order to introduce a complete frameworktoward a new philosophy of MCMC method and of statisticalinference method about Markov chains. First, we model clonalselection algorithms using Markov chains. Second, we carryon a particle analysis and analyze the convergence propertiesof these algorithms. Third, we propose the uni¯ed MCMC the-orem and unique chromosomes method for a purely successfuloptimization of these algorithms.