解非线性最优化问题的信赖域方法-计算数学专业论文.docxVIP

摘 摘 要 非线性最优化是研究非线性决策问题并寻求其最优解的最佳选择。信赖 域方法是求解非线性最优化问题的一类有效方法，它的优点是思想新颖，具有 较好的收敛性和可靠性。Levenberg-Marquardt方法是一种可以克服由雅可比 矩阵奇异或接近奇异所引起困难的优化方法。本论文结合了传统信赖域方法 和Levenberg-Marquardt方法的优点，给出了解非线性优化问题 砌min。愀剖2 的一种改进的信赖域方法。全文共分三章。 在第一章中，介绍了信赖域方法的背景和常见的信赖域方法，以及求解信 赖域子问题的算法。 在第二章中，我们提出了一种改进的信赖域算法。该算法的每次迭代中试 探步由信赖域子问题获得，信赖域半径用连续函数的方法更新，这个连续函数 以实际减少量和预计减少量的比率为自变量。改进的算法选取连续函数的方法 可以保证得到更大的信赖域半径，并且获得较好的数值结果。接下来讨论了改 进的信赖域算法具有传统信赖域算法的全局收敛性，在比非奇异性弱的局部误 差有界的条件下，由新算法产生的序列超线性收敛于方程组的解。 在第三章中，我们对比了改进的信赖域算法和以前的信赖域算法，数值例 子显示了改进的信赖域算法的优势，在迭代步数和残差上都要优于与其对比的 算法。 关键词：非线性方程组，最优化，信赖域，超线性收敛 AbstractNonlinear Abstract Nonlinear optimization is widely preferred in studying the nonlinear decision problem and find its optimal solution．The trust region method is an effective method to solve nonlinear optimization．It is a new idea and has good con- vergence．The advantage of Levenberg-Marquardt method is to overcome the difficulties caused by singularity or near singularity of the Jacobi．In this paper， we combines the advantage of the trust region and Levenberg-Marquardt method to consider a new improved trust region method for the nonlinear optimization m—in IIF(x)l[2． z∈Rn。 The whole article contains of three chapters． In the first chapter，we introduce the background of the trust region method． Several classical trust region methods and the algorithm for solving subproblem are reviewed． We introduce our improved trust region method in the second chapter．In each iteration，a trial step is obtained by solving the trust region subproblem．In the new algorithm we update trust region radius by a continuous function and Can obtain the larger trust region radius．The independent variable of the continuous function is the ratio of the actual reduction and the predicted reduction．It preserves the global convergence of the traditional trust region algorithm．Under the local error bound condition which is weaker than the nonsingularity,the sequence generated by the new algorithm converges superlinearly to the solution． I