TMLNNs：三值／多值逻辑神经元网络

TMLNNS:THREE OR MANY VALUED LOGIC NEURAL NETWORKS

本文提出了具有三值／多值逻辑表达能力的神经元模型（ＴＭＬＮ），即三值／多值＂逻辑与＂神经元（ＴＭＬＮ－ＡＮＤ）和三值／多值＂逻辑或＂神经元（ＴＭＬＮ－ＯＲ）．由这种逻辑神经元连接而成的多层神经网络（ＴＭＬＮＮｓ）能够实现三值／多值逻辑推理系统．本文还给出了ＴＭＬＮＮｓ的学习算法．从ＴＭＬＮＮｓ网络中容易抽取到三值／多值逻辑规则知识，可以将ＴＭＬＮＮｓ用于三值／多值逻辑规则知识的自动获取．ＴＭＬＮＮｓ模型为神经网络表示逻辑知识提供了理论基础．

A new neuron model(three or many valued logic neuron, or TMLN) is presented in this paper. It can express three valued logic knowledge and even many valued logic knowledge. There are two different TMLNs which are TMLN-AND (three or many valued 'logic and' neuron) and TMLN-OR (three or many valued 'logic or' neuron). Multilayer neural networks (TMLNNs) made up of this kind of neurons can be used as three or many valued logic inference system. The learning algorithm of TMLNNs is presented. Three or many valued logic rules can be extracted from TMLNNs easily. TMLNNs is the base for expressing logic knowledge using neural networks.