采用MCU多层神经网络模型设计探讨  

A Novel Model Design Adopting MCU for Multilayer Neural Network

作　　者：朱明程[1] 徐健[2] 李昆华[1] 

机构地区：[1]深圳大学信息工程学院,深圳518060 [2]中国科学院半导体研究所,北京100083

出　　处：《东南大学学报（自然科学版）》2000年第2期21-26,共6页Journal of Southeast University：Natural Science Edition

摘　　要：介绍了多层神经网络模型及BP学习算法 ,并以解决XOR问题为例 ,系统讨论了采用MOTOROLA单片机MC6 8HC11A8的多层神经网络模型的硬件设计与程序流程 ,进一步讨论了硬件设计中量化误差、算法的非线性等对系统性能的影响 。Multilayer neural network model and BP algorithms are introduced firstly in this paper. By solving XOR problem, hardware and software design of a novel multilayer neural network model adopting MC68HC11A8 of MOTOROLA MCU are dissertated systematically. The influence of quantizing errors in hardware design and the nonlinear algorithms to the system function is further discussed and the correction are adopted.

关 键 词：多层神经网络 BP算法 XOR问题 单片机 MCU 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]