Estimation Of Distribution Algorithm Based On Boosting

Estimation of Distribution Algorithm(EDA) is an evolutionary algorithm which generates new population by estimating the density of the old population to guide the population evolution. In this article we proposed a new EDA based on Boosting (BEDA), a methodology from statistical learning theory. We treat Boosting as a gradient descent procedure to minimize the risk in the space induced by weak learners and use it in EDA. To implement our algorithm we choose Gaussian distribution as weak learners for Boosting. Numerical results show BEDA out perform UMDA in some difficult function optimization problems.