RBF Surrogate-Assisted Optimization Algorithm For Large-scale Expensive Problems

With the complexity of engineering or scientific problems increases,the dimensionality of the optimization problems gradually increases.It takes a long time to evaluate one solution using the objective function for some problems.Since evolutionary optimization algorithms require a large number of objective function evaluations before obtaining the optimal solution or close to the optimal solution,it cannot be directly applied to solve large-scale expensive optimization problems.Surrogate-assisted evolutionary optimization algorithm is a common method to solve expensive optimization problems in recent years.However,the higher the dimension,the more training data will be required,which is difficult to obtain for expensive optimization problems.To this end,this paper uses random feature selection and random grouping to decompose large-scale problems into several low-dimensional sub-problems and constructs one surrogate model for each sub-problem,and then uses them to assist sub-problems optimization to solve large-scale expensive problems.The main contributions as follows:(1)A surrogate-assisted evolutionary algorithm with random feature selection(SAEA-RFS).A random feature selection technique is utilized to select decision variables from the original large-scale optimization problem to form a number of sub-problems,whose dimension may differ to each other,at each generation.The population employed to optimize the original large-scale optimization problem is updated by sequentially optimizing each sub-problem assisted by a surrogate constructed for this sub-problem.Then a new candidate solution of the original problem is generated by replacing the decision variables of the best solution found so far with those of the sub-problem that has achieved the best approximated fitness among all sub-problems.This new solution is then evaluated using the original expensive problem and used to update the best solution.In order to evaluate the performance of the proposed method,we conduct the experiments on 15 CEC'2013 benchmark problems and compare to some state-of-the-art algorithms.The experimental results show that the proposed method is more effective than the state-of-the-art algorithms,especially on problems that are partially separable or non-separable.(2)Pre-screening strategy assisted evolutionary algorithm(PSEA).In this algorithm,we divide a large number of decision variables into several non-overlapping sub-components by random grouping in cooperative co-evolutionary algorithm,thus forming several sub-problems,construct model on each sub-problem,then optimize and pre-screen them by the surrogate assisted evolutionary algorithm.Finally,by randomly splicing all the sub-populations,the original problem population is formed,all individuals in the population are evaluated using the expensive objective function,and save them into archive,which will be used to update training samples and global best solution.We have conducted a number of numerical experiments on CEC'2013,and compared with some state-of-the-art algorithms.Experimental results have demonstrated that the performance of this method is effective for solving large-scale optimization problems,especially on problems that are partially additively separable.