利用人工神经网络方法对牙形石鉴定中若干“模糊特征”的数字化研究  

Identification of the 'Ambiguity Features' of the Conodont by the Artificial Neural Network

作　　者：黄铮[1] 白志强[1] 柴华[1] 

机构地区：[1]北京大学地球与空间科学学院,北京100871

出　　处：《地质科技情报》2009年第3期94-98,共5页Geological Science and Technology Information

基　　金：教育部博士点基金项目(200800010052)

摘　　要：牙形石是一类在地层学研究中具有重要意义的古生物化石。但由于牙形石鉴定中存在诸多重要、却难以定量分析的"模糊特征",使得牙形石的鉴定和分类工作长期依赖于古生物工作者的经验。利用数学中的非线性方程将这些"模糊特征"进行数学描述,并利用人工神经网络中的反向传播算法设计出一套多隐含层的前馈网络分类器,对这些"模糊特征"进行分类尝试,取得了良好的效果,为牙形石的进一步数值化研究奠定了基础。Conodont is an important kind of fossil in the stratigraphy research. But ever since a long time ago the identification of the conodont has depended on the experience of the paleontologists because many of their features are significant but hard to be identified quantitatively. This paper describes these features with non-linear method in mathematics and designs a set of multiple hidden layer feed forward network based on the back propagation algorithm to identify those ＇ambiguity features＇. The method has yielded good effect, and laid a foundation for the further numerical research on the conodont.

关 键 词：牙形石 模糊特征 人工神经网络 反向传播算法 

分 类 号：Q911.23[天文地球—古生物学与地层学]