Comparison of performance of basic MEC and DC niching GAs

Mind evolutionary computation (MEC) proposed by Chengyi Sun (1998) is a new approach of evolutionary computation (EC). It has excellent performances on various aspects. In this paper we analyze the factors that influence deceptive degree of a function and build a series of functions to test different algorithms. First, the computing cost and search efficiency are defined. Then measurements of search efficiency and convergence rate are given to compare the searching performance of algorithms. Generally, the search efficiency of basic MEC is higher by above 40% than that of simple GA (SGA), especially for strongly deceptive problems, superiority of MEC is quite obvious. Compared with the search efficiency of DC (deterministic crowding) niching GA, the search efficiency of MEC is more than 50% higher. Also, the convergence ability of MEC is 70% higher than that of SGA, and over 50% than that of DC for most test functions.