Improvement And Research Of Yin Yang Pair Optimization Algorithm

Optimization problems have been always a key task in computer engineering area.Nowadays,more and more complex optimization problems are not suitable by applying classical optimization techniques.Intelligent optimization algorithms have been widley applied to solve multiple kinds of optimization problems,due to their high performance,high flexibility,and robustness.Currentlym researchers have proposed a banch of high performance intelligent optimizationa lgorithms,such as particle swarm algorithm,genetic algorithm,whale algorithm,etc.The yin-yang pair optimization algorithm(YYPO)is a recently proposed high performance but light weight algorithm.However,YYPO has some defects,such as being easily trapped into local optimum,relatively lower precision,and uneven distribution of initial solutions.In this paper,3 enhanced and improved YYPO algorthims are proposed and shown as follow:(1): In order to overcome the defect of being easily trapped into local optimum,an improved yin-yang pair optimization algorithm that adopts simulated annealing strategy is proposed,named YYPO-SA.YYPO-SA applies simulated annealing(SA)strategy in the switch phase between P1 and P2,and adopts adapted temperature adjustment mechanism to achive an appropriate temperature drop,and eventually help YYPO-SA to jump out of local optimals.the improved algorithm has a stronger ability to jump out of the local optimum and higher convergence accuracy.The experimental results show that YYPO-SA can significantly achieve better performance than YYPO.(2): In order to futher improve accuracy of YYPO-SA,an enhanced yin-yang pair optimization algorithm is proposed,named N-YYPO.N-YYPO adopts dynamical adjustment of D-direction segmentation probability,namely N-YYPO uses a higher D-direction segmentation probability during early iteration statge and gradually drops down the probability with iteration.NYYPO also applies a chaotic perturbation strategy to enrich the diversity of initial solutions.The experimental results show that the improved algorithm can achieve higher convergence accuracy and stronger global search ability than YYPO,especially in high-dimensional test functions.(3): An improvd algorithm,named YYPO-PT,is proposed,which introduces the dimension learning strategy and reverse search meanchism to further enhance the search effiency and final convergence precision of the algorithm.The two mechanisms introduced in YYPO-PT improve the D-spliting mechanism used in the original YYPO algorithm,since the D-spliting mechanism does not take the direction information of the already known optimal position into consideration when generating news points during each iteration.The experimental results illustrate that YYPOTP can also achieve better experimental results than YYPO-SA and N-YYPO algorithm.