自组织特征映射神经网络权值的实现

The Implementation of Synaptic Weight of Self Organizing Neural Network for Mapping Character

本文根据自组织特征映射神经网络学习算法，提出了其权值的ＣＭＯＳ实现电路。该电路结构简单，动态范围宽，易于集成实现。学习速度可达５０万次每秒。用ＣＭＯＳ１．２μ工艺实现电路的ＰＳＰＩＣＥ５．０仿真，表明该电路具有较好的工作性能。

According to a learning algorithm of self organizing neural network for mapping character, a CMOS implementation of its synaptic weight by circuit is presented in this paper.The circuit with a simple structure and a wide dynamic range is novel and easy to be realized. The learning rate is more than 5*10 4 times every second. The simulation of the circuit implemented with CMOS1.2μ can be verified validly by PSPICE5.0.