Research On Efficient Algorithms For Solving Large Scale Global Optimization Problems

With the development of science and technology,large-scale global optimization problems(LSGO)are increasingly used in scientific research and engineering applications.It is difficult to obtain global optimal solutions under limited computational resources due to its complex and large-scale search spaces,excessive local minima points,etc.Cooperative Coevolution(CC)algorithm is a kind of efficient algorithm for solving LSGO problems.It decomposes high-dimensional problem into a series of small-scale sub-problems with some grouping strategies,which significantly reduces the dimension of the problem and improves the optimization efficiency of the algorithm.However,the optimization efficiency of CC depends on the quality of the algorithm used to solve sub-problems and the grouping strategy.In order to improve the performance of CC,the main work of this article is as follows:(1)For the existing optimization algorithms used to solve sub-problem,they do not have high enough performance and are easy to fall into local optimal solutions when dealing with LSGO problems.Aiming at these problems,an algorithm called Multi Population based Selfadaptive Differential Evolution(MPSa DE)is proposed in this article.Firstly,a novel adaptive mutation strategy is designed to improve the efficiency of algorithm,which combines various mutation operators and assigns appropriate mutation operators to each individual according to the optimizing information.Besides,it randomly divides the population into multiple sub-populations,which reduces the influence range of the current optimal solution and thus the population can maintain a good diversity.Then,the parameters are dynamically adjusted to improve the efficiency of the algorithm.Since different mutation operators have different properties,it is unreasonable to use the same parameters for all mutation operators.Therefore,the parameters of different mutation operators here adaptively adjust separately,making full use of the characteristics of various mutation operators.Finally,a diversity detection mechanism is introduced into the algorithm to avoid premature convergence,which can enhance the population diversity appropriately according to the historical convergence of population and the current population information.MPSa DE is incorporated into the CC framework,and the test results on the common benchmark suits show that it has competitive performance.(2)There are many issues in the collaborative co-evolutionary algorithms now,such as low efficiency in dealing with large-scale sub-problems,low grouping efficiency,and unreasonable use of computing resources.In order to solve these issues,a new Cooperative Coevolution with Self-adaptive Resource Allocation and Hybrid Grouping(CC-Sa RA-HG)is proposed in this article.Firstly,the corresponding contribution-based grouping strategy is proposed for the fully separable and non-separable problems.Next,a grouping strategy with high accuracy is improved,which improves the efficiency in the grouping stage,so that more computing resources can be allocated to the optimization stage than before;Then,an adaptive regrouping strategy is proposed for the large-scale sub-problems that may exist after grouping.By comparing and selecting appropriate re-grouping strategies,the ability of the algorithm to deal with such problems is improved.Finally,in order to make full use of the limited computing resources,an intelligent resource allocation strategy suitable for the designed algorithm here is proposed.Compared with some state-of-the-art LSGO algorithms on different benchmarks suits,the algorithm proposed in this article has promising performance in experiments.