对自组织神经网络模式识别方法中若干问题的研究  

A study of some problems in pattern recognition by self-organizing neural network

作　　者：张登辉[1] 徐仲达[1] 

机构地区：[1]同济大学,上海200092

出　　处：《石油物探》1998年第2期103-111,70,共10页Geophysical Prospecting For Petroleum

摘　　要：模式识别方法已成为地震资料解释及储层预测的一个重要手段。在诸多的模式识别方法中，神经网络以其较强的智能性、准确性和容错性为人们所重视，应用十分广泛。本文在总结前人工作的基础上，用模型分析研究了自组织神经网络模式识别中应引起重视的几个问题，如参数的选取及适用性，噪声干扰对识别结果的影响等。这些结论在实际资料处理中起到了较大的指导作用。The pattern recognition has been an important means in seismic data interpretation and reservoirprediction. Among the pattern recognition methods, the neural network method has been widely used because of its powerful mtelligence, accuracy and allowance. Based on the surmmary of predecessors works, we study some problems in pattern recognition by self-organizing neural network via model analy-sis which should be paid attention to, for example, the selection and adaptation of parameters, and the effect of noise disturbance on the recognition results. These conclusions play a directive role in real dataprocessing.

关 键 词：自组织神经网络 石油勘探 模式识别 地震勘探 

分 类 号：TE19[石油与天然气工程—油气勘探] P631.4[天文地球—地质矿产勘探]