An Analysis of Estimation of Distribution Algorithms with Finite Population Models

The convergence of estimation of distribution algorithms (EDAs) with finite population is analyzed in this paper. At first, the models of EDAs with finite population are designed by incorporating an error into expected distribution of parent population. Then the convergence of the EDAs is proved with finite population under three widely used selection schemes. The results show that EDAs converge to the optimal solutions within the range of error described in this paper.