Research On Many-objective Evolutionary Algorithm Based On Adaptive Decomposition

In practical engineering and scientific research,it is necessary to optimize multiple objectives simultaneously,which are these objectives are often conflicting.In order to solve these problems,the traditional multi-objective evolutionary algorithm based on decomposition has shown strong performance.But there are still some disadvantages for many-objective optimization problems.Aiming at the above problems,different kinds of many-objective evolutionary algorithm based on adaptive decomposition have been put forward to optimize these problems.The main works of this thesis are as follows:(1)Considering the imbalance on convergence and diversity of the population which caused by raditional PBI decomposition algorithm,an adaptive decomposition evolutionary algorithm based on environmental information for many-objective optimization(Ma OEA/ADEI)is proposed.In this algorithm,an environmental selection strategy based on adaptive PBI decomposition is designed to select the optimal solution of the population.And the penalty factor of PBI is adaptively adjusted according to the environmental information.In order to solve the optimization problem with different scale objecives,the adaptive adjustment strategy of reference vector is put forward.The historical solution information in the process of population evolution is used in this method.Based on the simulation of DTLZ test problem,the proposed algorithm obtains a set of solutions with good convergence and diversity.(2)Aiming at the many-objective optimization problems with complex Pareto front,a many-objective evolutionary algorithm based on adaptive boosting learning(Ma OEA-ABL)is proposed.The adaptive boosting learning algorithm is designed to adjust the predefined reference vector.The associated characteristic of the reference vector and individuals are used as the evidence to divide reference vector into positive vector and negative vector.The adaptive boosting learning algorithm is used to training classfier for identify effective area of reference vector.In order to deal with many-objective optimization prolems with different curvature,an unbiased decomposition method of Pareto shape is designed.The decomposition strategy decompodes the objective space through the adaptive adjusted reference vector,so Ma OEA-ABL is a kind of adaptive decomposition algorithm.Simulation has been conducted on Ma F test problems.The experimental results show that the proposed algorithm Ma OEA-ABL performs well in many-objective optimization problems with complex Pareto front.