tuto8_sol_3230

pdf

School

The Chinese University of Hong Kong *

*We aren’t endorsed by this school

Course

3230

Subject

Mathematics

Date

Nov 24, 2024

Type

pdf

Pages

7

Uploaded by MinisterGorilla8297

Report
MATH3230 Numerical Analysis Tutorial 8 with solution 1 Recall: 1. LU factorization with partial pivoting: Idea: Permute the rows so that the largest entry in magnitude in the column becomes the pivot. Strategy: At the k th stage of the LU factorization, suppose the matrix A becomes A k = ( a ( k ) ij ) . Then determine an index p k for which | a ( k ) p k ,k | is the largest among all | a ( k ) j,k | for k j n . Then interchange rows k and p k before proceeding the next step of the factorization. 2 Exercises: Please do the star problems (*) in the tutorial class and finish the rest after the class. 1. (a) *Find the LU factorization of the given matrix. A = ஀? ஀? ஀? ஀? ஀? ஀? 1 2 2 3 5 2 6 6 10 12 2 6 9 22 21 3 10 22 69 63 5 12 21 63 75 ஀? ஀? ஀? ஀? ஀? ஀? (b) *Convert the LU factorization of the symmetric positive definite matrix A to the LDU factorization and U T U factorization. Solution. (a) Let L 1 = ஀? ஀? ஀? ஀? ஀? ஀? 1 0 0 0 0 2 1 0 0 0 2 0 1 0 0 3 0 0 1 0 5 0 0 0 1 ஀? ஀? ஀? ஀? ஀? ஀? , then L 1 A = ஀? ஀? ஀? ஀? ஀? ஀? 1 2 2 3 5 0 2 2 4 2 0 2 5 16 11 0 4 16 60 48 0 2 11 48 50 ஀? ஀? ஀? ஀? ஀? ஀? L 2 = ஀? ஀? ஀? ஀? ஀? ஀? 1 0 0 0 0 0 1 0 0 0 0 1 1 0 0 0 2 0 1 0 0 1 0 0 1 ஀? ஀? ஀? ஀? ஀? ஀? , then L 2 L 1 A = ஀? ஀? ஀? ஀? ஀? ஀? 1 2 2 3 5 0 2 2 4 2 0 0 3 12 9 0 0 12 52 44 0 0 9 44 48 ஀? ஀? ஀? ஀? ஀? ஀? L 3 = ஀? ஀? ஀? ஀? ஀? ஀? 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 4 1 0 0 0 3 0 1 ஀? ஀? ஀? ஀? ஀? ஀? , then L 3 L 2 L 1 A = ஀? ஀? ஀? ஀? ஀? ஀? 1 2 2 3 5 0 2 2 4 2 0 0 3 12 9 0 0 0 4 8 0 0 0 8 21 ஀? ஀? ஀? ஀? ஀? ஀? L 4 = ஀? ஀? ஀? ஀? ஀? ஀? 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 2 1 ஀? ஀? ஀? ஀? ஀? ஀? , then L 4 L 3 L 2 L 1 A = ஀? ஀? ஀? ஀? ஀? ஀? 1 2 2 3 5 0 2 2 4 2 0 0 3 12 9 0 0 0 4 8 0 0 0 0 5 ஀? ஀? ஀? ஀? ஀? ஀? = U 1
Let L = ( L 4 L 3 L 2 L 1 ) 1 = ஀? ஀? ஀? ஀? ஀? ஀? 1 0 0 0 0 2 1 0 0 0 2 1 1 0 0 3 2 4 1 0 5 1 3 2 1 ஀? ஀? ஀? ஀? ஀? ஀? Then A = LU = ஀? ஀? ஀? ஀? ஀? ஀? 1 0 0 0 0 2 1 0 0 0 2 1 1 0 0 3 2 4 1 0 5 1 3 2 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 1 2 2 3 5 0 2 2 4 2 0 0 3 12 9 0 0 0 4 8 0 0 0 0 5 ஀? ஀? ஀? ஀? ஀? ஀? (b) By (a), we have in LDU : D = ஀? ஀? ஀? ஀? ஀? ஀? 1 0 0 0 0 0 2 0 0 0 0 0 3 0 0 0 0 0 4 0 0 0 0 0 5 ஀? ஀? ஀? ஀? ஀? ஀? and U = ஀? ஀? ஀? ஀? ஀? ஀? 1 2 2 3 5 0 1 1 2 1 0 0 1 4 3 0 0 0 1 2 0 0 0 0 1 ஀? ஀? ஀? ஀? ஀? ஀? And in U T U : U = ஀? ஀? ஀? ஀? ஀? ஀? 1 2 2 3 5 0 2 2 2 2 2 0 0 3 4 3 3 3 0 0 0 2 4 0 0 0 0 5 ஀? ஀? ஀? ஀? ஀? ஀? 2. Let A be nonsingular and the LU factorization with pivoting takes the form: U = L n 1 P n 1 · · · L 2 P 2 L 1 P 1 A, where L i are elementary matrices, U is an upper triangular matrix and P i are permutation matrices for pivoting. (a) *Explain when it is necessary to do the LU factorization with partial pivoting. (b) Write down the idea of the LU factorization with partial pivoting. Do this factorization for the following matrix A and find each of the matrix U , L i , P i for i = 1 , 2 , · · · , n 1 , where n is the row number of A . A = ஀? ஀? ஀? ஀? ஀? ஀? 3 6 2 5 1 5 5 2 4 4 6 6 8 2 2 4 7 9 3 8 5 3 5 7 2 ஀? ஀? ஀? ஀? ஀? ஀? . (c) Denote P = P n 1 · · · P 1 . Prove that PA = LU , where L is a lower triangular matrix with 1 as its diagonal entries and all entries of L satisfy that | l ij | 1 . (d) Find P , L , U in (c) according to question (b) such that PA = LU , where A is the matrix in (b). (e) Use the decomposition in (d) to solve the system Ax = b , where A is the one in (b) and b = (46 , 57 , 60 , 97 , 64) T . Solution. 2
(a) When the absolute value of diagonal entry is too small at certain stage of the LU factorization, it may force algorithm to stop or cause bu ff er overflow. (b) R 3 R 1 ஀? ஀? ஀? ஀? ஀? ஀? 1 1 1 1 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 3 6 2 5 1 5 5 2 4 4 6 6 8 2 2 4 7 9 3 8 5 3 5 7 2 ஀? ஀? ஀? ஀? ஀? ஀? = ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 5 5 2 4 4 3 6 2 5 1 4 7 9 3 8 5 3 5 7 2 ஀? ஀? ஀? ஀? ஀? ஀? LU ஀? ஀? ஀? ஀? ஀? ஀? 1 5 6 1 1 2 1 2 3 1 5 6 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 5 5 2 4 4 3 6 2 5 1 4 7 9 3 8 5 3 5 7 2 ஀? ஀? ஀? ஀? ஀? ஀? = ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 0 14 3 7 3 7 3 3 2 4 0 3 11 3 5 3 20 3 2 5 3 16 3 1 3 ஀? ஀? ஀? ஀? ஀? ஀? R 3 R 2 ஀? ஀? ஀? ஀? ஀? ஀? 1 1 1 1 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 0 14 3 7 3 7 3 3 2 4 0 3 11 3 5 3 20 3 2 5 3 16 3 1 3 ஀? ஀? ஀? ஀? ஀? ஀? = ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 0 14 3 7 3 7 3 3 11 3 5 3 20 3 2 5 3 16 3 1 3 ஀? ஀? ஀? ஀? ஀? ஀? LU ஀? ஀? ஀? ஀? ஀? ஀? 1 1 0 1 1 1 2 3 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 0 14 3 7 3 7 3 3 11 3 5 3 20 3 2 5 3 16 3 1 3 ஀? ஀? ஀? ஀? ஀? ஀? = ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 14 3 7 3 7 3 17 3 7 3 20 3 3 8 1 3 ஀? ஀? ஀? ஀? ஀? ஀? R 4 R 3 ஀? ஀? ஀? ஀? ஀? ஀? 1 1 1 1 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 14 3 7 3 7 3 17 3 7 3 20 3 3 8 1 3 ஀? ஀? ஀? ஀? ஀? ஀? = ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 17 3 7 3 20 3 14 3 7 3 7 3 3 8 1 3 ஀? ஀? ஀? ஀? ஀? ஀? LU ஀? ஀? ஀? ஀? ஀? ஀? 1 1 1 14 17 1 9 17 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 17 3 7 3 20 3 14 3 7 3 7 3 3 8 1 3 ஀? ஀? ஀? ஀? ஀? ஀? = ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 17 3 7 3 20 3 3 17 133 17 115 17 197 51 ஀? ஀? ஀? ஀? ஀? ஀? R 5 R 4 ஀? ஀? ஀? ஀? ஀? ஀? 1 1 1 1 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 17 3 7 3 20 3 3 17 133 17 115 17 197 51 ஀? ஀? ஀? ஀? ஀? ஀? = ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 17 3 7 3 20 3 115 17 197 51 3 17 133 17 ஀? ஀? ஀? ஀? ஀? ஀? LU ஀? ஀? ஀? ஀? ஀? ஀? 1 1 1 1 7 115 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 17 3 7 3 20 3 115 17 197 51 3 17 133 17 ஀? ஀? ஀? ஀? ஀? ஀? = ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 17 3 7 3 20 3 115 17 197 51 2618 345 ஀? ஀? ஀? ஀? ஀? ஀? 3
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
  • Access to all documents
  • Unlimited textbook solutions
  • 24/7 expert homework help
(c) From pivoting form: U = L n 1 P n 1 ...L 2 P 2 L 1 P 1 A Now we claim that there exists some L ( k ) such that L n 1 P n 1 ...L 2 P 2 L 1 P 1 = L ( n 1) L ( n 2) . . . L (1) P n 1 P n 2 . . . P 1 such that the structure of L ( k ) is equal to L k but with the subdiagonal entries permuted. We prove the claim as follows: Define L ( n 1) = L n 1 , L ( n 2) = P n 1 L n 2 P 1 n 1 , L ( n 3) = P n 1 P n 2 L n 3 P 1 n 2 P 1 n 1 , . . . we have L ( n 1) L ( n 2) . . . L (1) P n 1 P n 2 . . . P 1 = L n 1 ( P n 1 L n 2 P 1 n 1 )( P n 1 P n 2 L n 3 P 1 n 2 P 1 n 1 ) . . . P n 1 P n 2 . . . P 1 = L n 1 P n 1 L n 2 P n 2 . . . L 1 P 1 (1) Therefore we have U = ( L ( n 1) L ( n 2) . . . L (1) )( P n 1 P n 2 . . . P 1 ) A As L ( k ) has the same structure as L k for all k , we can combine them to form a lower triangular matrix with 1’s as diagonal entries. Therefore we have proved the result. (d) L = ஀? ஀? ஀? ஀? ஀? ஀? 1 1 2 1 2 3 1 1 5 6 2 3 9 17 1 5 6 0 14 17 7 115 1 ஀? ஀? ஀? ஀? ஀? ஀? , U = ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 17 3 7 3 20 3 115 17 197 51 2618 345 ஀? ஀? ஀? ஀? ஀? ஀? , P = ஀? ஀? ஀? ஀? ஀? ஀? 1 1 1 1 1 ஀? ஀? ஀? ஀? ஀? ஀? (e) We use Ax = b to denote the system in (a). Applying PA = LU and now we solve LUx = Pb. Righthand-side is Pb = ஀? ஀? ஀? ஀? ஀? ஀? 1 1 1 1 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 46 57 60 97 64 ஀? ஀? ஀? ஀? ஀? ஀? = ஀? ஀? ஀? ஀? ஀? ஀? 60 46 97 64 57 ஀? ஀? ஀? ஀? ஀? ஀? and lefthand-side LU = ஀? ஀? ஀? ஀? ஀? ஀? 1 1 2 1 2 3 1 1 5 6 2 3 9 17 1 5 6 0 14 17 7 115 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 6 6 8 2 2 3 2 4 0 17 3 7 3 20 3 115 17 197 51 2618 345 ஀? ஀? ஀? ஀? ஀? ஀? Then we solve x = U 1 L 1 Pb = U 1 ஀? ஀? ஀? ஀? ஀? ஀? 1 1 2 1 2 3 1 1 5 6 2 3 9 17 1 5 6 0 14 17 7 115 1 ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? ஀? 60 46 97 64 57 ஀? ஀? ஀? ஀? ஀? ஀? = U 1 ஀? ஀? ஀? ஀? ஀? ஀? 60 16 41 2365 / 51 2618 / 69 ஀? ஀? ஀? ஀? ஀? ஀? = ஀? ஀? ஀? ஀? ஀? ஀? 1 2 3 4 5 ஀? ஀? ஀? ஀? ஀? ஀? 4
3. Let A R n × n be strictly column diagonally dominant, i.e., there holds | a kk | > ஀? j = k | a jk | , k = 1 , · · · , n. (a) *Prove that a strictly diagonally dominant matrix is nonsingular. (b) Prove that if the strictly diagonally dominant matrix is symmetric and with positive diagonal entries, then it is positive definite. (c) Show that if LU factorization with partial pivoting is applied to A , no row interchanges take place. Solution. (a) We can prove by contradiction. Suppose A is singular, then A T is also singular. There exists x = 0 such that A T x = 0 . Let k { 1 , 2 , ..., n } be such that | x k | = max 1 j n {| x j |} > 0 Since A T x = 0 , we have n ஀? j =1 a jk x j = 0 a kk x k = ஀? j = k a jk x j Take absolute value on both sides, we have | a kk x k | = ஀? ஀? ஀? ஀? ஀? ஀? ஀? j = k a jk x j ஀? ஀? ஀? ஀? ஀? ஀? | a kk | | x k | ஀? j = k | a jk | | x j | | a kk | | x k | ஀? j = k | a jk | | x k | | a kk | < | a kk | Contradiction occurs. So A is non-singular. (b) We can prove it by contradiction. Suppose A has a negative eigenvalue λ and let u be the corresponding eigenvector. Note that λ and u are nonzero. Let k { 1 , 2 , ..., n } be such that | u k | = max 1 j n | u j | > 0 Then we have Au = λ u ( a kk λ ) u k = ஀? j = k a kj u j Take absolute value on both sides, we have | a kk λ | | u k | ஀? j = k | a kj | | u j | | a kk λ | | u k | ஀? j = k | a kj | | u k | | a kk λ | ஀? j = k | a kj | | a kk λ | < | a kk | 5
Contradiction occurs. (c) Since | a 11 | > ஀? j =1 | a j 1 | , there is no need to interchange row for the first step. Then we do the Gaussian elimination. We denote the A (1) be the matrix A after the first step Gaussian elimination. Then we obtain for j > 1 , a (1) ij = a ij a i 1 a 11 a 1 j . Hence, k > 1 , | a (1) kk | = | a kk a k 1 a 11 a 1 k | | a kk | | a k 1 a 11 || a 1 k | > ஀? i = k | a ik | | a k 1 a 11 || a 1 k | = ஀? i> 1 ,i = k a (1) ik (2) Therefore A (1) is also strictly diagonally dominant matrix. The argument repeats and the result is proved. 4. In this question, we want to prove that: If A is nonsingular, then there exist permutation matrices P 1 and P 2 , a unit lower triangular matrix L , and a nonsingular upper triangular matrix U such that P 1 AP 2 = LU. Note that if A is nonsingular, then it has a nonzero entry. So we can choose permutation matrices P 1 and P 2 such that the (1,1)-th position of P 1 AP 2 is nonzero. (a) Let P 1 AP 2 = ஀? a 11 A 12 A 21 A 22 ஀? = ஀? 1 0 L 21 I ஀? · ஀? u 11 U 12 0 ஀? A 22 ஀? . Show that ஀? A 22 is nonsingular. (b) Suppose there exists ஀? P 1 and ஀? P 2 such that ஀? P 1 ஀? A 22 ஀? P 2 = ஀? L ஀? U, where ஀? L is a unit lower triangular matrix and ஀? U is a nonsingular upper triangular matrix. Show that there exits permutation matrices P 1 and P 2 , such that P 1 AP 2 = ஀? 1 0 ஀? P 1 L 21 ஀? L ஀? ஀? u 11 U 12 ஀? P 2 0 ஀? U ஀? . (c) Using (a) and (b) with inductive argument, prove the theorem stated above. Solution. (a) Since det( P 1 AP 2 ) = ± det( A ) = 0 and also det( P 1 AP 2 ) = det ஀? 1 0 L 21 I ஀? · det ஀? u 11 U 12 0 ஀? A 22 ஀? = u 11 · det( ஀? A 22 ) Moreover P 1 AP 2 = ஀? a 11 A 12 A 21 A 22 ஀? = ஀? 1 0 L 21 I ஀? · ஀? u 11 U 12 0 ஀? A 22 ஀? = ஀? u 11 U 12 L 21 u 11 L 21 U 12 + ஀? A 22 ஀? Therefore we have a 11 = u 11 = 0 . So ஀? A 22 is invertible. 6
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
  • Access to all documents
  • Unlimited textbook solutions
  • 24/7 expert homework help
(b) Note P 1 AP 2 = ஀? 1 0 L 21 I ஀? ஀? u 11 U 12 0 ஀? P T 1 ஀? L ஀? U ஀? P T 2 ஀? = ஀? 1 0 L 21 I ஀? ஀? 1 0 0 ஀? P T 1 ஀? L ஀? ஀? u 11 U 12 0 ஀? U ஀? P T 2 ஀? = ஀? 1 0 L 21 ஀? P T 1 ஀? L ஀? ஀? u 11 U 12 ஀? P 2 0 ஀? U ஀? ஀? 1 0 0 ஀? P T 2 ஀? = ஀? 1 0 0 ஀? P T 1 ஀? ஀? 1 0 ஀? P 1 L 21 ஀? L ஀? ஀? u 11 U 12 ஀? P 2 0 ஀? U ஀? ஀? 1 0 0 ஀? P T 2 ஀? so we get a desired factorization of A : P 1 AP 2 = ஀?஀? 1 0 0 ஀? P 1 ஀? P 1 ஀? A ஀? P 2 ஀? 1 0 0 ஀? P 2 ஀?஀? = ஀? 1 0 ஀? P 1 L 21 ஀? L ஀? ஀? u 11 U 12 ஀? P 2 0 ஀? U ஀? (c) Using (a) and (b) and repeat the argument on the submatrix. The statement is proved. 7