A Study Of Particle Swarm Optimization Based On Inexact One-Dimensional Search And Adaptive Moment Estimation

The concept of particle swarm optimization algorithm(PSO)relative to other swarm intelligence algorithm is simple,easy to implement,need to adjust less parameters.Therefore,it has been widely applied.Although PSO has better global search capability,it has large blindness in population search,which reduces convergence accuracy and speed.Due to adaptive moment algorithm has quick local convergence and ideal overall convergence under certain conditions,the adaptive moment estimation algorithm is used to further local search based on the results of the basic particle swarm optimization algorithm and then the search ability of particles in the search space is improved;Because of the characteristics of directional search and local convergence,the inexact one-dimensional search algorithm is introduced to speed up the convergence rate of the algorithm.The main work of thesis is as follows:(1)A particle swarm optimization algorithm based on adaptive moment estimation(AdamPSO)is proposed to solve the problem that the adaptive particle swarm optimization algorithm(APSO)can not jump out of the local optimum in time and has lower convergence accuracy.On the basis of APSO search,the adaptive moment estimation is used to local search for each particle.According to the fitness value of the particle,the search direction of the particle is adjusted to find a better solution.The experimental results of the algorithm in multimodal test function show that the proposed algorithm is significantly improved and requires less iterative times compared with the basic PSO algorithm and the same type of algorithm based on gradient.(2)A particle swarm optimization algorithm based on inexact one-dimensional search and adaptive moment estimation(ILS-AdamPSO)is proposed to solve the problem that the search time is increased due to the repetitive search phenomenon in the initial stage of the basic particle swarm optimization algorithm.The inexact searchalgorithm is first used for deterministic search until the diversity of population is lost,and then the diversity of the population is increased according to the improved velocity evolution equation.Finally,when the global optimal value of the population has not been changed after several iterations,the adaptive moment estimation algorithm is used to update the global optimum of the population.The experimental results show that,compared with other algorithms,the new algorithm not only improves the searching ability of the particle swarm in the search space,but also makes the particles trapped in the local optimal point to jump out and move towards a better solution when solving the optimal value of the multimodal test function.