一种新的元胞网络模型  

A New Kind of Cellular Network

作　　者：李学伟[1] 孙有发[2] 吴今培[3] 

机构地区：[1]大连交通大学,辽宁大连116028 [2]广东工业大学管理学院,广东广州510520 [3]五邑大学智能技术与系统研究所,广东江门529020

出　　处：《五邑大学学报（自然科学版）》2011年第4期14-21,共8页Journal of Wuyi University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(No.70571006;No.70801019)

摘　　要：用元胞替换传统人工神经网络中的神经元,以局部连接取代相邻层级元胞之间的全连接,用规则演化算法替代BP算法,建立元胞网络模型.设计了元胞网络的训练过程:内嵌的遗传算法用于寻优各元胞状态离散化边界值以及元胞网络拓扑结构.以一个红酒分类基准数据为例,检验了元胞网络的学习性能和分类性能,获得了较为满意的结果.A novel concept of cellular network is proposed. The cellular network is constructed based on the traditional artificial neural network by replacing the neurons with ceils, replacing the full connections among cells with partial connections, and replacing the back-propagation algorithm with rule-evolution algorithm. The training process of cellular network is also designed. The embedded genetic algorithm is to search for the optimal parameters and structures of the network. Finally, the cellular network is applied to a benchmark data set （classification of Italy＇s red wine）, and is validated to be of good performances of learning and predicting.

关 键 词：元胞自动机 神经网络 遗传算法 规则演化 元胞网络 

分 类 号：TP393.01[自动化与计算机技术—计算机应用技术]