基于分组混沌PSO算法的模糊神经网络建模研究

发布时间：2018-04-25 23:23

本文选题：混沌思想 + 分组并行粒子群算法　；参考：《计算机工程与应用》2017年09期

【摘要】：为改善记忆功放建模的精度,且针对粒子群算法早期收敛速度较快,但在后期易陷入早熟收敛,局部最优等特点,提出了一种分组并行混沌粒子群优化算法(Grouping Parallel-Chaotic Particle Swarm Optimization,GPCPSO),将分组粒子群优化算法与混沌思想相结合,并用该算法优化动态模糊神经网络(Dynamic Fuzzy Neural Network,DFNN)参数,建立DFNN功放模型。引入分组的CPSO群算法,将种群划分为若干个组,每组单独计算,大大提高了收敛速度,同时将混沌思想运用到每个粒子当中去,避免早熟和局部最优,缩短了迭代时间。通过仿真结果可以看到,GP-CPSO优化后的动态模糊神经网络建模的训练误差减小到0.1以内,收敛速度提高32.5%,从而验证了这种建模方法有效且可靠。
[Abstract]:In order to improve the modeling accuracy of memory amplifier, the particle swarm optimization (PSO) algorithm can converge quickly in the early stage, but it is easy to fall into premature convergence and local optimum in the later stage. A group parallel chaotic particle swarm optimization (PSO) algorithm called grouping Parallel-Chaotic Particle Swarm Optimization is proposed, which combines the PSO algorithm with the chaotic idea. The algorithm is used to optimize the parameters of dynamic Fuzzy Neural Network (DFNN) and establish the DFNN power amplifier model. The CPSO swarm algorithm of grouping is introduced to divide the population into several groups, each group is calculated separately, which greatly improves the convergence rate. Meanwhile, chaos is applied to each particle to avoid precocity and local optimization, and the iterative time is shortened. The simulation results show that the training error of GP-CPSO optimized dynamic fuzzy neural network model is reduced to less than 0.1, and the convergence speed is increased by 32.5, which proves that this modeling method is effective and reliable.
【作者单位】：辽宁工程技术大学电子与信息工程学院;
【基金】：国家自然科学基金(No.61372058) 辽宁省高等优秀人才支持计划项目(No.LR2013012) 辽宁省教育厅科学研究一般项目(No.L2015209)
【分类号】：TP18

【相似文献】