摘要
本文改进了传统神经网络进行学习后权值矩阵在工作时不变的缺陷,给出一种处理多对单模式分类问题新思路:网络工作时,权值矩阵中的特定局部的取值可变。并通过此思路给出一种新型神经网络模型──局部可变权值神经网络。
A new idea is presented in this paper to model the default that weighted matrix isn't variable after study finishing: during network working, the special local values of weighted matrix may be variable. Finally a new neural network model-local variable weighted neural network is given.
出处
《浙江海洋学院学报(自然科学版)》
CAS
1999年第4期305-309,共5页
Journal of Zhejiang Ocean University(Natural Science Edition)