【作者】 田库； 王俊松； 刘玉敏； 谢超；

【Author】 TIAN Ku, WANG Jun-song, Liu Yu-min, Xie Chao(Tianjin University of Technology and Education, Tianjin 300222)

【机构】 天津职业技术师范学院自动化教研室；

【摘要】 本文基于离散泰勒级数提出一种对任意阶多维函数可实现无差逼近的新型联想记忆系统——DTS-AMS,在此基础上设计了一种基于DTS-AMS的交通流预测方案,详细讨论了其基本原理、插值算法及训练规则。基于DTS-AMS的交通流预测方案相对于CMAC-AMS具有学习精度高、收敛速度快、所需内存单元少、无杂散编码引起的信息丢失等优点。且比多层BP神经网络学习算法简单、计算量小。仿真实例表明了系统的可行性与有效性。更多还原

【Abstract】 The on-line intelligent rolling predictive method of traffic flow is proposed based on a novel high-order Associative Memory System, which is designed via Discrete Taylor Series(DTS-AMS), and is capable of implementing error-free approximations to multi-variable polynomial functions of arbitrary order. The advantages of the predictive model based on DTS-AMS offers over that based on conventional CMAC-type neural network are: high-precision of learning, much smaller memory requirement without the data-collision problem and also the advantages of much less computational effort for training and faster convergence rates than that attainable with multi-layer BP neural networks. The simulation results show that the predictive method is effective. The future work is the hardware realization of the predictive algorithm.更多还原