论文部分内容阅读
本文提出一种新的非线性反演方法——随机共轭梯度法。该方法采用非启发式反演方法 ,快速收敛到某一极值 ;再用启发式反演方法跳出局部极值 ;然后使用非启发式反演方法收敛到另一局部极值 ,反复进行此过程 ;并在解空间范围内搜索 ,保留所有的局部极值 ,最终确定最优解。它继承了随机爬山能够全局寻优、共轭梯度法计算速度快和精度高的优点 ,能快速搜索到全局最优解。试验证明 ,这种方法是一种高效的反演算法 ,特别适用于求解非线性、多极值的最优化问题和地球物理反问题。
In this paper, a new nonlinear inversion method - stochastic conjugate gradient method is proposed. The method uses a non-heuristic inversion method to quickly converge to an extreme value; then heuristic inversion method is used to jump out of the local extremum; and then the non-heuristic inversion method converges to another local extremum and the process is repeated ; And searching within the solution space, retaining all the local extremums, finally determining the optimal solution. It inherits the advantages of global optimization of random climbing, conjugate gradient method of fast calculation and high precision, and can quickly search the global optimal solution. Experiments show that this method is an efficient inversion algorithm and is especially suitable for solving nonlinear and multi-value optimization problems and geophysical inverse problems.