基于改进蚁群算法的DRNN网络动态系统辨识研究
首发时间:2009-10-09
摘要:针对BP算法训练对角递归神经网络(DRNN)时的缺点及传统蚁群算法收敛速度慢的问题,提出了一种改进的蚁群优化算法。该算法在传统蚁群算法(ACA)的基础上,借鉴粒子群算法(PSO)中粒子最优位置转移机制,采用一种新颖的动态信息素更新策略,并进行遗传操作策略来加快局部寻优。实验仿真结果表明,与基于BP算法和基于传统蚁群算法的DRNN网络模型相比较,基于改进蚁群算法的DRNN网络新模型的性能,尤其是网络的泛化能力,收敛速度和辨识精度都得到了较大程度的提升。
For information in English, please click here
DRNN Based on improved ant colony algorithm and Its Application to Dynamical System Identification
Abstract:For the shortcomings of BP Algorithm for training on the Diagonal Recurrent Neural Network (DRNN) and traditional Ant Colony Algorithm for the problem of slow convergence, proposed a new improved Ant Colony Optimization Algorithm. On the basis of the traditional Ant Colony Algorithm, learning Particle Swarm Optimization (PSO) in the optimal position of particle transfer mechanisms,using a novel Dynamic Pheromone update strategy and conduct genetic manipulation strategies to speed up the local optimization.The results show the new optimization DRNN model based on improved Ant Colony Optimization Algorithm is superior to that of other two kinds of optimized DRNN models,and the network has been a greater degree of upgrading especially in generalization ability,convergence rate and identification accuracy.
Keywords: DRNN network System indentification Particle swarm optimization Optimization
论文图表:
引用
No.3568031957612550****
同行评议
共计0人参与
勘误表
基于改进蚁群算法的DRNN网络动态系统辨识研究
评论
全部评论0/1000