摘要
Ｂｏｏｌｅ函数的线性可分和线性不可分问题，一直是前向人工神经网络的一个比较困难的问题，目前仅对变量数ｎ＜７的线性可分问题给予过讨论。本文基于简化问题讨论和优化问题的解的思想，提出了稳健分类复杂度的概念，并对Ｂｏｏｌｅ函数稳健分类复杂度为１的稳健线性可分问题，提出了对这类Ｂｏｏｌｅ函数进行计数的方法。
The linear and non linear separability of n dimensional Boolean functions is one of the most difficult problems in discrete feedforward neural networks,where the linear separability of n <7 dimensions was only discussed before. On the basis of simplifing the problems and optimizing the solutions of the problems, the concepts of the tolerant linear separability and its complexity of n dimensional Boolean functions are presented in this paper,with the result presented of the numbering of Boolean functions with their tolerant linear separability complexity to be 1.
出处
《系统工程与电子技术》
EI
CSCD
1999年第1期72-78,共7页
Systems Engineering and Electronics
基金
国家自然科学基金
电科院预研基金
关键词
神经计算电路
布尔函数
稳定性
人工神经网络
Feedforward neural networks,Linear and non linear separability,Clustering complexity and tolerant clustering complexity, N dimensional hypercube, N dimensional, Boolean function.
