Continuous optimization based-on greedy estimation of GMM

A new estimation of distribution algorithms (EDAs) for continuous optimization based on greedy estimation of Gaussian mixture mode is proposed. By incrementally adding new components one by one, the new estimation method can approximate almost any complex probability density function efficiently, so it has the ability to learn the model structure and parameters automatically without any requirement for prior knowledge. Since for each estimation iteration the task is simplified to be a two-component mixture model learning problem, and a greedy strategy is adopted to guarantee the monotonous increasing of likelihood, the new EDA is very fast and efficient. A set of experiments has been implemented to evaluate, and to compare with other EDAs, the efficiency and performance of the new algorithm. The results show that, with a relative small number of generations, the new algorithm can perform very well on both uni-modal and multi-modal function optimization problems