一种基于多层感知器的机器人快速推理学习方法  

A Illative Network for Robot Learning Based on the Multi-layer Perception

作　　者：宋宜斌[1] 王培进[1] 李凯里[1] 

机构地区：[1]烟台大学计算机学院,山东烟台264005

出　　处：《计算机工程与应用》2004年第13期57-59,71,共4页Computer Engineering and Applications

基　　金：山东省自然科学基金项目(编号:Y2002G04)

摘　　要：从多层感知器原理分析出发,该文提出一种适变学习因子法用于对学习算法的改进,并将改进的算法用于“逃避”机器人推理网络的实例样本的学习。仿真结果表明,改进后BP的算法可显著加速网络训练速度,并且学习过程具有较好的收敛性及较强的鲁棒性.According to the principle of Multilayer Perception(MLP),this paper presents a method with adaptive learn-ing rate factors for the improvement of MLP learning algorithm.The improved algorithm is applied to the learning of a illative network for the escaping process of Robot.The simulations show that the improved algorithm has good effects on speeding up learning process and makes its learning convergence better and robust performance stronger.

关 键 词：多层感知器 BP算法 推理网络 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]