论文部分内容阅读
为提高量子行为粒子群算法的优化能力,提出了一种改进的算法.该算法也采用量子势阱作为寻优机制,但提出了新的势阱中心建立方法.在每步迭代中,首先计算粒子适应度,然后取前K个适应度最好的粒子作为候选集.采用轮盘赌策略在候选集中选择一个粒子作为势阱中心,调整其它粒子向势阱中心移动.在优化过程中,通过使K值单调下降,获得探索与开发的平衡.将提出的算法应用于标准函数极值优化和量子衍生神经网络权值优化,实验结果表明提出算法的优化能力比原算法确有明显提高.
In order to improve the optimization ability of quantum behavior PSO, an improved algorithm is proposed, which uses quantum well as the optimization mechanism, but proposes a new method of establishing trap center.In each iteration, we first calculate Particle fitness, and then take the first K best fitness particles as a candidate set.Use roulette strategy in the candidate set to select a particle as a potential well center, adjust the other particles to move to the potential well center.In the optimization process, through So that the value of K decreases monotonously to obtain the balance between exploration and development.The proposed algorithm is applied to the optimization of standard functions and the weight of quantum-derived neural networks.The experimental results show that the proposed algorithm has better optimization ability than the original algorithm.