A New Approach to Generate Solutions Combining Crossover and Estimation of Distribution Operators for EMO Algorithm

Most of EMO algorithms use crossover operator for generating new solutions. There have been proposed various kinds of crossover in this field and most crossover approaches are good at global optimization but not effective for the problem with strong nonlinearity and dependency. On the other hand, Estimation of Distribution Algorithm (EDA) is known as an effective approach without using crossover for generating new solutions. EDA use an estimation of distribution operator for generating new solutions and this operator is known as to be effective for the problem with strong nonlinearity and dependency. In this paper, a new approach to generate new solutions combing crossover and estimation of distribution is proposed. The main purpose of this approach is to generate high-quality solutions more effectively by combing each other's strength. This approach is named as "MOEA/D Combined with Estimation of Distribution (MOEA/D-CED)" because this approach is incorporated with MOEA/D. Through applying to some benchmark problems in this field, the characteristics and effectiveness of MOEA/D-CED were confirmed by the comparison with original MOEA/D and MO-CMA-ES.