论文部分内容阅读
Occam反演法由于其算法稳定,对初始条件要求不高,反演效果较好,在大地电磁反演中运用较多。不过其每次迭代都需要进行偏导数计算和大量的模型计算,以便搜索到最佳的拉格朗日乘子,这造成了计算量和计算时间的增加。Occam反演在每次迭代寻找最佳模型的过程中需要搜索合适的拉格朗日乘子使拟合差最小,搜索的方法一般使用进退法和扫描法,鲜少使用其他一维搜索方法。本文将牛顿迭代搜索法和二分法组合一起用于拉格朗日乘子的搜索,取得了较好的结果,减少了模型的搜索量,在一定程度上提高了计算速度。
Due to its stable algorithm, the Occam inversion method does not require much initial conditions, and the inversion effect is better, so it is more used in the magnetotelluric inversion. However, each iteration requires partial derivative calculations and a large number of model calculations in order to search for the best Lagrange multipliers, which results in an increase in computation and computation time. Occam inversion needs to search the appropriate Lagrange multipliers to minimize the fitting error during the search for the best model for each iteration. The search method generally uses the forward and backward method and the scanning method, and seldom uses other one-dimensional search methods. In this paper, we use Newton’s iterative search method and dichotomy combination together for the search of Lagrange multipliers, and achieved good results, reducing the search volume of the model and improving the computing speed to a certain extent.