一种Q学习的量子粒子群优化方法

摘要

分析了量子行为粒子群优化算法，着重研究了算法中的收缩扩张参数及其控制方法，针对不同的参数控制策略对算法性能的影响特点，提出将Q学习方法用于算法的参数控制策略，在算法搜索过程中能够自适应调整选择参数，提高算法的整体优化性能；并将改进后的Q学习量子粒子群算法与固定参数选择策略，线性下降参数控制策略和非线性下降参数控制策略方法通过CEC2005 benchmark测试函数进行了比较，对结果进行了分析。%Quantum-behaved Particle Swarm Optimization algorithm is analyzed; contraction-expansion coefficient and its control method are studied. To the different performance characteristics with different coefficients control strategies, a control method of coefficient with Q-learning is proposed. The proposed method can tune the coefficient adaptively and the whole optimization performance is increased. The comparison and analysis of results with the proposed method, con-stant coefficient control method, linear decreased coefficient control method and non-linear decreased coefficient control method based on CEC2005 benchmark function is given.