非光滑方程的算法及其利用.pdf

摘 要 非光滑方程问题是与求解互补问题、变分不等式问题、工程力学问题、金融 分析等诸多问题密切相关的一类优化问题，非光滑方程问题的研究包含了理论研究 与算法设计两个方面。 本文第一章考虑了非光滑方程中的极大值方程问题的求解算法，给出了求解 极大值方程的最速下降法和光滑梯度法。此外，给出了算法的收敛性分析及在求 解广义互补问题、极大极小优化问题中的应用。最后的光滑梯度算法的数值试验 表明了算法在实际应用中的有效性。 本文第二章继续对在经济管理、工程技术等领域有广泛应用的极大极小非光 滑问题进行了研究，给出了光滑的Fletcher-Reeves 共轭梯度算法。在一般条件下， 证明了算法的全局收敛性，给出了相应的数值实验。 本文的最后一部分，给出了由Facchinei F 等人提出的一种关于求解约束非光 滑方程的牛顿型算法，该算法主要利用线性规划模型来求解约束非光滑方程问题。 结合算法特征，给出了算法在求解约束极大值方程和约束广义互补问题中的应用。 关键词：非光滑方程；全局收敛；极大极小问题；光滑梯度法 万方数据 Abstract Nonsmooth equations problem is a class of optimization problems which is closely related to the complementarity problems, variational inequality problem, engineering mechanics, financial analysis and many other optimization problems. The study of nonsmooth equations problems include theory research and algorithm design. The first chapter of this paper considers solving the nonsmooth equations with finitely many maximum functions, gives the steepest descent method and smoothing gradient method to solve the nonsmooth equations. In addition, we also give the convergence analysis of the method and the application in solving the generalized complementarity problem and the minimax optimization problem. Finally, the numerical experiments show that the smoothing gradient method is effectiveness in practical application. This second chapter of this paper continues to study the minimax nonsmooth problem which is widely applied in the fields of economic management and engineering te