基于牛顿力学和博弈论模型的粒子网络优化算法

发布时间：2018-01-11 10:08

本文关键词：基于牛顿力学和博弈论模型的粒子网络优化算法　出处：《山东大学学报(工学版)》2017年01期 　论文类型：期刊论文

【摘要】：为克服标准粒子群算法在求解高维TSP问题时求解精度不高、易陷入局部最优等不足,将每个粒子均赋予质量和加速度,利用泊松分布和牛顿第二运动定律动态调整粒子加速度,并将粒子维数以相似度划分为优势部分和劣势部分,正常更新时只对劣势部分进行相应处理,保持并扩大其优势部分以提高收敛速度,扰动时更新其优势部分以达到远离当前粒子网络的目的来跳出局部最优。当有粒子碰撞时,引入反向学习策略处理粒子,选择合适的降速模型来提高收敛速度。最后,将改进后的算法用于求解TSPLIB中的标准实例问题,并与经典算法进行比较。试验结果表明,提出的新算法在求解旅行商问题时具有高效率、低迭代次数及强收敛等特性。该结果可为智能算法在求解优化问题时提高精确性和加快收敛等方面的研究提供一定的参考。
[Abstract]:In order to overcome the shortcomings of standard particle swarm optimization (PSO) in solving high dimensional TSP problem with low accuracy and easy to fall into local optimum, each particle is given mass and acceleration. The particle acceleration is adjusted dynamically by using Poisson distribution and Newton's second law of motion, and the particle dimension is divided into the advantage part and the inferior part by similarity, and only the inferior part is dealt with when updating normally. The dominant part is preserved and expanded to improve the convergence rate, and the dominant part is updated when disturbed to reach the goal of escaping from the current particle network to jump out of the local optimum. When there is a particle collision, the local optimum can be obtained. The reverse learning strategy is introduced to deal with particles, and the appropriate deceleration model is selected to improve the convergence rate. Finally, the improved algorithm is used to solve the standard instance problem in TSPLIB. The experimental results show that the proposed algorithm is efficient in solving the traveling salesman problem. The results can provide a certain reference for the research of improving the accuracy and speeding up the convergence of the intelligent algorithm in solving the optimization problem.
【作者单位】：河池学院计算机与信息工程学院;河池学院智能计算与模式识别重点实验室;武汉大学计算机学院;河池市人民政府办公室;
【基金】：国家自然科学基金资助项目(61170305) 河池学院科研启动经费资助项目(XJ2016KQ01) 国家级大学生创新创业训练计划资助项目(201610605029)
【分类号】：TP18
【正文快照】： 0引言标准粒子群优化算法[1-2]简单且易实现,参数设置少,受到学术界广泛重视,但算法也存在易陷入局部最优、收敛速度慢等缺陷。研究自然计算的专家学者们对传统粒子群算法进行改进。其中,SHIY和EBERHART[3]提出基于模糊系统的惯性权重的动态调整,从而实现对惯性权重的非线性控

【参考文献】