四元数神经网络模型：彩色图象的识别  

THE HAMILTON NEURAL NETWORK MODEL RECOGNITION OF THE COLOR PATTERNS

作　　者：帅建伟[1] 陈振湘[1] 刘瑞堂 吴伯僖[1] 

机构地区：[1]厦门大学物理系

出　　处：《计算机学报》1995年第5期372-379,共8页Chinese Journal of Computers

基　　金：国家自然科学基金

摘　　要：本文首次把四元数（Ｈａｍｉｌｔｏｎ量）引入到神经网络中，提出了四元数１６态离散神经网络模型，应用信噪比理论和计算机模拟考察了该模型的存储稳定性和存储容量该模型的存储容量比与Ｈｏｐｆｉｅｌｄ模型相同．该四元数神经网络模型可应用于１６色的彩色图象识别中．In this paper, the Hamilton nubber is introduced into the neural network model, and the discrete Hamilton need network model is suggested,whereby a neuron is a 16-state (±1±i±j±k). The signal-to-noise theory and the computer simulation are made to analyse the storage stability and the storage capacity of the model. The storage eapacity ratio of the model equals to that of the Hopfield model. The 16-state Hamilton number neural network can be applied to recognize the 16-level color patterns.

关 键 词：神经网络 四元数 图象识别 彩色图象 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]