摘要
为使接收函数的反演更为简便,本文提出了一种基于人工神经网络误差反传(BP)算法的接收函数反演新方法,该方法采用人工神经网络反演系统,避免了接收函数反演过程中复杂的地震响应计算及耗时的雅可比矩阵计算,只需经过学习训练就能够解决复杂的实际问题,而且具有记忆功能,这使接收函数的反演工作具有延续性和可继承性.理论数据的反演计算结果表明,该方法是切实可行的.
A new inversion method of artificial neural network is presented for the purpose of simplification and more general application for the inversion method in receiver functions research. On the basis of back propagation（BP） algorithm, an adaptive inversion system of artificial neural network had been designed and developed. The system avoids complicated calculations of seismic response and Jacobi matrix. It also can be used to solve some practical problem through learning and training. This method can memorize and make the inversion a continuous process which depends not only on current event but also the previous event.
出处
《地球物理学进展》
CSCD
北大核心
2007年第6期1916-1920,共5页
Progress in Geophysics
基金
中国科技部与美国科学基金(USNSF)共同资助的课题(国土资源部合同编号:2001010206)资助
关键词
接收函数
人工神经网络
BP算法
速度结构反演
receiver functions, artificial neural net work, BP algorithm, inversion of velocity
