Estimation of Distribution Algorithms Based on Mixtures of Factor Analyzers

The estimation of distribution algorithms(EDAs) based on mixtures of factor analyzers is proposed.The selected optimal individuals are clustered with rival penalized competitive learning algorithm,and for each single cluster,the factor analyzer model is used to estimate its distribution information.In order to maintain the population diversity,the algorithm retains the individuals which have better fitness and farther away from the sets of the selected individuals,and uses the parameters of clustering to reduce computation cost.Experimental result approves the performance of the algorithm.