The Improvement Research Of Differential Algorithm

In recent years, the evolutionary computation as a new rapidly developing subject, it is a simulation of biological evolution and evolutionary mechanisms for solving the problem of self-organizing, adaptive artificial intelligence technology, differential evolution algorithm is a branch of evolutionary algorithms. In 1995 Rainer Storn and Kenneth Price proposed evolution algorithm, which is an optimization iterative calculation method based on population. Since the differential evolution algorithm is simple structure, good convergence performance, control less parameters, easy to implement and to receive widespread attention, and has been used function optimization, pattern recognition, artificial neural networks, combinatorial optimization problems.This paper analyzes the differential evolution algorithm research status and improvement strategies, given the theoretical evolutionary algorithm is extremely implementation techniques, followed by the basic principle of differential evolution algorithm, initialization method populations, variation equations and parameters of performance were analyzed. Although evolutionary algorithms have many advantages, but there are also easy to fall into local optimum, late convergence rate is not high, easy precocious and other shortcomings. In this paper, dynamic local search differential evolution algorithm and cuckoo search differential evolution algorithm are proposed to improve differential evolution algorithm.The idea of dynamic local research differential evolution algorithm is: random choice method and small probability perturbation are adopted to increase the diversity of population and balance exploitation and exploration. The method makes full use of dynamic local research to optimize the current best solution to speed up the convergence rate. Experiments are conducted on a suite of benchmark functions and the results are compared with those of other algorithms. The results demonstrate the proposed algorithm has a faster convergence rate and higher solution accuracy, and shows good performance in solving complex numerical optimization problems.The idea of cuckoo search differential evolution algorithm is: A new mutation strategy based on the optimal solution guided equation is adopted. For the individual which is not updated in the selection stage, we use cuckoo search strategy to produce a new individual to select again. In order to speed up the convergence rate, the method makes full use of dynamic local research to optimize the current best solution. Experiments are conducted on a suite of benchmark functions and the results are compared with those of other five algorithms. The proposed algorithm shows good performance in solving complex numerical optimization problems.