基于复数权值神经元的数字逻辑最稳健设计  

Robust design of digital logic based on neurons with complex-valued weights

作　　者：吕伟锋[1] 林弥[1] 华柏兴[1] 

机构地区：[1]杭州电子科技大学电子信息学院

出　　处：《电路与系统学报》2008年第4期77-80,共4页Journal of Circuits and Systems

摘　　要：复数权值神经元由于引入了多阈值逻辑而具有更强的性能,文中根据其数学模型,结合二进神经元稳健性的概念,提出了基于该神经元的稳健性定义。并结合定义,实现了基于单个神经元的基本数字逻辑以及异或运算的最稳健设计方案,从而证明了该稳健神经元实现任意数字逻辑的有效性和可行性,说明了其强大的处理能力。Neurons with complex-valued weights have stronger capability because of its multi-valued threshold. In the paper, definition of robustness for this kind of neurons is proposed according to their mathematical model and binary neurons＇ conception of robustness. In addition, robust design for basic digital logic and XOR operations of two varieties are implemented based on the neuron. So it is believed the validity and possibility with the robust MVN for realizing arbitrary digital logical function as well as its strong process capability.

关 键 词：复数权值 多值神经元 数字逻辑 异或问题 稳健神经元 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TN431[自动化与计算机技术—控制科学与工程]