Co-evolutionary global optimization algorithm

A hybrid global optimization method, the coevolutionary global optimization algorithm, is proposed which utilizes the self-organized critical state as the mean of diversification of search and the traditional conjugate gradient local minimization method as the mean of intensification of search. The former has been recently used by Boettcher and Percus (Artificial Intelligence 119 (2000) 275) to solve discrete combinatorial optimization problems. The proposed method has been tested to locate the lowest energy conformation of atomic clusters. It was found that the method was effective not only to locate the lowest energy state but also to enumerate all the low-lying metastable states.