Research On The Cuckoo Search Algorithm For Multi-modal Function Optimization

Cuckoo search algorithm is a new heuristic algorithm,which is used to solve the global function optimization problem.Cuckoo search algorithm adopts Levy flight random walk model in the process of optimization.It is easy to implement and has a few number of control parameters,excellent search path and strong optimization capability,and has been successfully applied to practical problems,such as engineering optimization.There are two problems in Cuckoo search algorithm.The first problem is that the step produced by Levy flight is random,but it lacks self-adaptability and controllability.So it has the problem of slow convergence rate and unstable search accuracy in later stage.The second problem is that the original cuckoo search algorithm is a global search algorithm,which is only suitable for obtaining the global optimal solution,and can not solve the multi-modal function oriented problem to find multiple local optimal solutions.The main work of this paper is as follows.(1)In order to solve the problem of lack of controllability in the process of algorithm evolution,this thesis proposes a Cuckoo Search Algorithm based on Stochastic Gradient Descent.This algorithm uses Stochastic Gradient Descent in Levy flight random walk process to improve the controllability and self-adaptability in the evolutionary process and enhance the search of the local optimum,convergence process and algorithm adaptability,which improves the calculation accuracy and convergence rate of cuckoo search algorithm.Multiple simulation experiments show that the CS_SGD algorithm is simple and efficient,improves the performances on calculation accuracy and convergence rate on the basis of maintaining the advantages of the standard CS algorithm.(2)In order to solve the multi-modal function oriented problem to find multiple local optimal solutions,this thesis introduces the niche technology based on sharing mechanism in the Cuckoo Search Algorithm based on Stochastic Gradient Descent,which increases species diversity during algorithm evolution.Simulation experiments show that the improved algorithm has the advantages of fast convergence rate,high accuracy of searching the optimal solutions and small computational complexity in solving multi-modal function oriented problem,and also has broad practical application value.