Real-Coded Estimation of Distribution Algorithm

In this paper, we show how Estimation of Distribution Algorithms (EDAs) can be applied to the optimization of multivariate functions in continuous domain. First, we describe the relationship between Gaussian network model and multivariate normal densities along with the methods used for the learning and simulation of Gaussian networks. Next, we propose our algorithm that uses only the means and covariance matrix of the variables, estimated from the selected promising individuals of a population, to generate offspring. Finally, we apply that algorithm to three bench-mark functions: Summation cancellation, Ackley and Sphere model, and provide the experimental results. The experimental results show that our proposed algorithm may produce results more accurately and more efficiently than some other EDAs.