Research on compact genetic algorithm in continuous domain

Compact genetic algorithm (CGA) is a successful probability-based evolutionary algorithm which performs equivalent to the order-one behavior of the simple genetic algorithm (SGA) with uniform crossover. However, this equivalence only applies for binary encoded problems. To extend the basic concept of CGA to continuous domain, an improved CGA is proposed in this paper. We established a continuous CGA (cCGA) model by adopting two probability vectors to represent population. We study the update rules of the probability vectors and its initial value. In further we improve this cCGA by adopting elitism selection. We propose two kinds of elitism based cCGA by applying different elitism control policies. Theoretical analysis on elitism control is given and some useful results are concluded. The numerical experiment first gives a comparison between SGA and our cCGA in continuous domain and the results show the superiority and efficiency of cCGA. Comparison between elitism selection cCGA and non-elitism cCGA is also given to show the efficiency of elitism selection and the efficiency on elitism control.