地球物理资料非线性反演方法讲座(五) 人工神经网络反演法  

Lecture on Non-Linear Inverse Methods in Geophysical Data(5) The Artificial Neural Network Method

作　　者：王家映[1] 

机构地区：[1]中国地质大学地球物理与空间信息学院,武汉430074

出　　处：《工程地球物理学报》2008年第3期255-265,共11页Chinese Journal of Engineering Geophysics

摘　　要：木文概要地介绍了近年来发展起来的一种强有力的地球物理资料非线性反演的方法——人工神经元法(ANN)。比较系统地介绍了ANN的基本知识和两种基于不同原理的ANN反演方法。基于模式识别和基于最小方差的ANN反演方法。列举了这两种方法在地球物理资料反演中的一些应用实例、指出了该方法的存在问题和发展方向。In this paper we briefly introduced the most powerful nonlinear inverse method -- artificial neural network, which has been developed and used successfully in geo- physical data inversion in last two decades. We not only introduced some basic knowledge of ANN, but also analyzed two kinds of inverse methods, founded in publications, based on pattern recognition and minimum variance. We also described the application of the two methods in geophysical data. Finally, we pointed out the problem of the methods and future development.

关 键 词：非线性反演方法 人工神经元法 BP回传网络 HOPFIELD网络 

分 类 号：P631[天文地球—地质矿产勘探]