Research On Parallel And Optimization Of Gravitational Field Optimization Algorithm

Optimization theory and method is an important method used by people to select the best method from multiple solutions and approaches when solving practical problems.Optimization problems are common in human society.With the diversification and discretization of practical problems,the running time for solving events of combinatorial optimization problems show exponential growth.In order to achieve the optimal solution to practically complex problems,modern optimization algorithms emerged in the 1980 s.The Gravitational Field Algorithm is a modern optimization algorithm proposed by Zheng Ming,which mainly simulates celestial mechanics and is derived from mathematical modeling based on the Solar Nebular Disk Model(SNDM).It simulates the process of planet formation to search for the optimal solution.Although this optimization algorithm has more advantages than other optimization algorithms in multi-peak optimization problems,when the global optimal value has a higher accuracy requirement,the GFA needs to calculate a lot of initial dust,reducing the efficiency of the algorithm.Therefore,this paper proposes an optimization method based on multipopulation parallel(coarse-grained model)to accelerate the Gravitational Field Algorithm(GFA).With the help of MATLAB's computing platform and its parallel toolbox,the parallel computing method of multi-core CPUs is mainly used to implement PGFA,which improves the speed of algorithm execution.At the same time,this paper also improves the mobile operation and absorption operation in the parallel GFA.While improving the speed of algorithm execution,improve the accuracy of algorithm solution.Finally,by selecting the test functions of eight classical unconstrained optimization problems,the experimental tests are compared from the perspectives of variable dimensions,number of groups and initial scales.The experimental results indicate that the PGFA has a higher calculation efficiency than the original GFA,and can improve the accuracy of algorithm optimization.It also proves the effect of the number of parallel gravitational field algorithm grouping on the algorithm execution efficiency.PGFA can further improve the speed of the algorithm with the increase of the number of cores,and has good scalability.Compared with GFA,PGFA has more advantages in optimization problems with large initial scale and high complexity.