二进神经网络中汉明球的逻辑意义及一般判别方法  

LOGICAL MEANING OF HAMMING SPHERE AND ITS GENERAL JUDGEMENT METHOD IN BINARY NEURAL NETWORKS

作　　者：陆阳[1] 魏臻[1] 高隽[1] 韩江洪[1] 

机构地区：[1]合肥工业大学计算机与信息学院,合肥230009

出　　处：《计算机研究与发展》2002年第1期79-86,共8页Journal of Computer Research and Development

基　　金：国家自然科学基金资助 (6980 5 0 0 1)

摘　　要：剖析二进神经元的逻辑意义对二进神经网络的规则提取是十分重要的 .在布尔空间中 ,汉明球是一种线性可分的空间结构 ,如何从汉明球中提取出具有清晰逻辑意义的规则是二进神经网络研究中的一个问题 .通过对Mof N规则表达形式的扩展 ,分析了汉明球的逻辑意义 ,提出了表达汉明球逻辑意义的 L EM规则和 GEM规则方法 ,并且讨论了汉明球和汉明补球的等价性 .另一个重要的结果是证明了二进神经元和汉明球等价的充要条件 。For extracting rules from binary neural networks, it is important to analyze logical meaning of neurons. Hamming sphere is a kind of structure in Boolean space. How to extract rules which express clear logical meaning from Hamming sphere is a problem in binary neural network research. In this paper, clear logical meaning of Hamming sphere is dug out by modifying the MofN rule form to LEM and GEM rule forms, which are proposed as two new rule forms, and also the equivalent relation between a Hamming sphere and a Hamming complementary sphere is discussed. Another important result is that a necessary and sufficient condition for a Hamming sphere equivalent to a binary neuron is proved. By that necessary and sufficient condition, a method of judging a Hamming sphere is established.

关 键 词：二进神经网络 汉明球 规则提取 逻辑意义 机器学习 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]