几类非线性矩阵方程的迭代解法-数学专业论文.docxVIP

PAGE PAGE IV Abstract Abstract The problem of solving nonlinear matrix equation is one of important issuses in the field of numerical algebra and non-linear science in recent years. Actually, it is widely used in many fields such as structure design, system identification, dynamic programming, automatic control theory, vibration theory, statistics. In this paper, we use Newton method, Steepest descent method, conjugate gradient method, BB gradient method and some new kinds of methods to solve some kinds of nonlinear matrix equation. In Chapter 2, we study the existence of its positive definition solution, showing the existence of positive definition solution. Moreover the two methods were proposed to this nonlinear matrix equation, the convergence of the methods is also given. At last, the numerical experiments show the feasibility of our methods. In Chapter 3, we proposed a new method for the square root of matrix’ the constraint solution, the symmetric matrix solution and the central the symmetric matrix solution are researched, the convergence of solving symmetric matrix solution method is also given, the numerical examples were given to illustrate the feasibility of our algorithm. If we make a small change about some iteration steps, we can solve linear or other nonlinear matrix equation with some constrained solutions, Through several methods show the correct of our speculation. In Chapter 4, we mainly concern about the general solution of the general quadratic matrix equations. At first we show some common used methods, such as Newton’s method, Newton’s method with exact line searches, the conjugate gradient method and so on. At the end, the BB gradient method was proposed for the equation. A comparison with some existing methods was also given. Through some experiments we show that the BB gradient method time was better than other methods for some special structure. The time compared with other methods is less. Key words: Nonlinear matrix equation; positi