計算課程講義

1、精選優(yōu)質(zhì)文檔-傾情為你奉上非線性方程組求解 完成日期09-8-20or whereand迭代法尋找一個數(shù)列 使得 迭代步長 使得迭代終止 （收斂） 或者 超過最大迭代次數(shù) （發(fā)散）Algorithm given 初始化 , > 0，最大迭代次數(shù)maxrepeat1. 計算 和 或者 2. if , return .3. 計算或者4. until max牛頓迭代法 n=1 This is called the linearized equation.n>1 其中Unconstrained minimization無約束最優(yōu)問題 convex diffmin , 其中牛頓迭代法 其中

2、Algorithm given 初始化 , > 0，最大迭代次數(shù)maxrepeat1. 計算 和 2. if , return .3. 計算4. until maxEquality constrained minimization等式約束最優(yōu)問題 with ; convex diff, .min subject to 牛頓迭代法, .步長為當(dāng)為可行性解（）內(nèi)點法(primal-dual interior-point methods),KKT minimize subject to 其中 convex diff ; with .我們要解如下非線性方程組 所以 其中 The primal-d

3、ual search direction線性方程組求解 完成日期09-8-21 LU分解； Cholesky分解（） i.e, QR分解 i.e, QR分解；Cholesky分解Nonsingular matrix The LU factorizationWhere P is an -permutation matrix, L is a unit lower triangular-matrix ( a lower triangular matrix with diagonal elements equal to one),and U is nonsingular upper triangula

4、r -matrix.Positive definite matrix The Cholesky factorizationwhere L is lower triangular with positive diagonal elements matrix with zero nullspace The QR factorization Where Q is -orthogonal matrix and R is an -upper triangular matrix with positive diagonal elements.Triangular matrix：A matrix A is

5、upper triangular if if 。A triangular matrix is called nonsingular if all the diagonal elements are nonzero. It is singular if one or more of the diagonal elements are zero.Forward/Back substitutionA is a nonsingular lower triangular matrix of order n. are解法Flop count: 1+3+5+(2n-1)=n2 。Forward/Back s

6、ubstitution (Recursive formulation) then The Cholesky factorization: Positive definite matrix positive definite: A is symmetric and for all nonzero x. , where L is lower triangular with positive diagonal elements therefore and is lower triangular with positive diagonal elements.makes:Therefore , , .

7、 In fact, . So, , and that makes positive definite matrix. Schur Complement (of positive definite matrix elementis positive definite) whereis a nonzero n-1 vector then x is a nonzero n vector, so .Therefore is positive definite matrix of order n-1. Solving linear equations by Cholesky factorization為了計算,