【摘要】 地震数据反演是一个典型的非线性逆问题。其目标函数都是多极值的函数，因此，传统的迭代优化方法常常会遇到局部收敛性的限制。本文提出的同伦神经优化理论（HNOT）及其算法（HNOA）能将非线性多极值目标函数较快地收敛于全局极值，是一种有效的反演方法。本文将该反演方法与相邻道互相关技术和层位信息约束有机地结合起来，实现了地震数据控制下的井资料高分辨率岩性参数联合反演。理论模型与实际资料的处理结果表明，本文提出的方法是可行的。更多还原

【Abstract】 Seismic data inversion is a typical nonlinear inverse problem, and 1ts objectfunctions are multiextremal functions. Therefore l ordinary iterative optimizati0nmethod are often restricted by local convergence. Homotopic neural optimizationtheory and its algorithm described in this paper can make nonlinear multiextremalobject functions converge fast to global extremum, so it is an effective inversionmethod. This inversion method is combined with both adjacent trace crosscorre-lation and stratum data restriction to jointly derive high-resolution l1thology param-eters from logging data under restriction of seismic data- Theoretical and real re-sults prove this method feasible.更多还原