摘要
采用对称三值逻辑的数元｛1，0，1｝作为信息存储的基本单位，本文提出一种用于逼近非线性函数的多值学习网络（KLN）．该网络由多个既关联又独立的子网络构成，而每个子网络包含一个权值存储单元组和一个阈值存储单元组．所需的数学运算仅为整数的加法和逻辑判断，因而非常简单．在此基础上，研究了具有自学习功能的多值逻辑学习控制策略．仿真结果表明KLN对非线性函数具有良好的学习和表达能力，并对复杂非线性系统具有良好的学习控制性能，有计算简单、省时的特点．
In this paper, a multi-valued learning network, KLN, is proposed for approximating nonlinear functions, with the symmetric ternary element { 1, O, 1 }as its basic information storage unit. This network is constructed by a set of distributively associated subnets, each of them contains two groups of memory units, i. e.,weight and threshold, and the required mathematical operations are just integer addition and logic judgement. Based on the proposed network, the multi-valued logic control strategy with self-learning functions is further studied. Simulation results show that this network has strong learning capacity for representing nonlinear functions, has good learning control performance for complex nonlinear systems, and has characteristics of simplification in computation.
出处
《计算机学报》
EI
CSCD
北大核心
1998年第6期553-559,共7页
Chinese Journal of Computers
基金
国家自然科学基金
