Running-time Analysis And Estimation For Swarm Intelligence Algorithm

Swarm intelligence algorithm plays an important role of the field of artificial intelligence.It is widely used in various of optimization problems and has achieved good results.Compared with its rich application research results,the theoretical research of swarm intelligence algorithms is quite scarce,especially the study of computation time.However,the study of computational time allows people to better understand the algorithm,understand the parameters of the algorithm,and parameters setting.Particle Swarm Optimization(PSO)and Ant Colony Optimization(ACO)are two typical swarm intelligence algorithms,which are widely used in continuous optimization problems and discrete optimization problems respectively.Therefore,this paper will analyze the computation time of PSO and ACO by theoretical analysis.However,complex theoretical analysis is difficult to apply to computation time analysis of complex and efficient real algorithms.Therefore,after theoretical analysis,this paper proposes an experimental estimation method to estimate the computation-time of PSO and ACO by combining theory with experiments.This paper is mainly divided into the following four parts.In the first part,computation-time analysis of two single-particle PSOs is proposed.In this paper,the average gain model is used to calculate the average gain of the algorithm by calculating the probability distribution of each iteration of the algorithm,and then the computation time of the algorithm is calculated based on average gain model.The computation time of the two algorithms is further compared.In the section part,an existing estimation method is improved to estimate the computation time of two PSO variants.The existing estimation method can be used for the computationtime estimation of the continuous optimization algorithm,but the estimation result cannot show the influence of the algorithm parameters on the computation time.Therefore,in this paper the existing estimation method is improved.The estimation results can show the influence of algorithm parameters on the computation time.Through this method,the computation time of two PSO variants is estimated,and the parameters of these algorithms are discussed through the estimation results.In the third part,the computation time of two Ant System(AS)algorithms are compared.The pheromone matrix is ?a matrix that guides the search direction of the algorithm maintained during the iteration of the AS.In this paper,the computation time of the two algorithms are compared by analyzing the changes of the pheromone matrix.In the fourth part,a method for estimating discrete optimization problems is proposed and further used to estimate the computation time of several ACO variants.Based on the existing computational time estimation method of continuous optimization algorithm,this paper proposes an estimation method of discrete optimization algorithm,which is used to estimate the computation time of several ACO variants.