一维大地电磁Occam反演拉格朗日乘子的搜索

Lagrange multiplier's search in one-dimensional magnetotelluric Occam's inversion

在大地电磁反演中,Occam法因其在反演稳定性和模型分辨率等方面的优势,得到广泛应用。但由于其每次迭代都需要不断地搜索拉格朗日乘子,因而拉格朗日乘子的搜索效率对Occam法反演的运算速度起着至关重要的作用。为提高拉格朗日乘子的搜索效率,这里提出将拉格朗日乘子的搜索从自然数域中转到以10为底的对数域中进行,同时在区间最小值的搜索中采用二次函数极小值搜索法。通过大量一维大地电磁反演验证,该方法将拉格朗日乘子的搜索效率提高了20%~50%,大大提高了Occam反演的运算速度,方便了其在高维反演中的应用。

In magnetotelluric inversion,because of its stability in inversion and the advantage of model resolution and so on,Occam method is widely used.But in each iteration,Occam's inversion needs to constantly search Lagrange multiplier.The searching efficiency of Lagrange multiplier to Occam's inversion speed plays a vital role.To improve the search efficiency of Lagrange multiplier,a new searching method was presented in this paper.First Lagrange multiplier of transfer from the natural number field to search for the 10 logs base domain.Second in the search of interval minimum search method using quadratic function.Through a number of one-dimensional magnetotelluric inversions,Lagrange multiplier search efficiency is increased by 20%- 50%,and the operation speed of Occam's inversion is greatly improved.