逻辑神经网络非线性可分性的实现

Implementation of Nonlinear Separability for Logic Neural Nets

１９８５年，Ａｌｅｋｓａｎｄｅｒ领导的小组所实现的逻辑神经网络（ＬＮＮ）成功地应用于脸谱实时识别。由于ＬＮＮ无需学习算法，硬件可实现实时识别能力，在英国颇受工业界重视。然而，ＬＮＮ存在一个缺陷就是在大模式划分成小的子模式后会造成ＬＮＮ非线性能力下降甚至消失。针对上述问题，提出了一种在ＬＮＮ中实现非线性可分性的方法。

In 1985,the logic neural network(LNN)implemented by Aleksander heading the research group was applied to real-time face recognition and industrial part test successfully,LNN got great emphasis from the U. K. industry for its simplified learning process, its hardware implementability and real-time recognition ability. But its drawback is that its nonlinear separable ability tends to diminish when large input patterns are divided into small subpatterns. In this paper, the authors propose a method to overcome the drawback of LNN.