基于CUDA的边界变异量子粒子群优化算法

Boundary Mutation Quantum-behaved Particle Swarm Optimization Algorithm Based on CUDA

针对量子粒子群优化算法面对复杂优化问题时,临近最优解的搜索阶段存在收敛速度慢、在边界附近全局搜索性差的问题,提出了基于CUDA的边界变异量子粒子群优化算法.GPU(图形处理器)以多颗密集的计算核心模拟粒子的搜索过程,利用并发的优势提升粒子搜索速度;边界变异则通过以随机概率将边界粒子扩散到更大的搜索域,增加种群的多样性,提升粒子群的全局搜索性.对若干优化算法的仿真实验表明,所提出方法具有较好的全局收敛性,且同等目标精度下,取得了较高的有效加速比.

For quantum particle swarm optimization algorithm in the face of complex optimization problems, and the near optimal solution search stage has the problem of slow convergence speed and poor global search, a new quantum particle swarm optimization based on CUDA is proposed in this paper. The multiple dense core was used to simulate particle search process, and the advantages of concurrent was used to improve the speed of particle search. The boundary variation is to increase the diversity of the population and improve the global search of the particle swarm by using the random probability. The simulation results of several optimization algorithms show that the proposed method has better global convergence performance. And with the same target accuracy, the high effective speedup ratio is obtained.