【摘要】 以Matlab平台为基础,利用神经网络工具箱构建了自组织神经网络,对已知沉积相的安徽宿南等矿区的19组样本进行SOM分类,并与系统聚类分类结果进行比较;指出在无监督分类粒度分析中,SOM方法分类操作过程简便易行,具有残缺自动识别能力,分类结果惟一,可以在沉积物成因分类中应用。更多还原

【Abstract】 A self-organizing mapping(SOM) artificial neural network is created based on the neural net toolbox of Matlab and used to classify 19 soil samples the sedimentary types of which has been recognized.The results are compared with those obtained by using the method of hierarchical clustering,and it is concluded that the SOM network can be applied conveniently to non-supervisor classification,and that it can identify incomplete samples without any prior knowledge and lead to a unique result.The good effect of classification proves that the SOM network can be applied to grain-size analysis.更多还原