Introduction to Linear algebra-Strang-Solutions-Manual_ver13

# Tb 12 a a d 1 1 has a a d m a b d b1 c c bm

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: The column space is the line through a D 31 The error e D b p must be perpendicular to all the a’s. 32 Since P1 b is in C .A/; P2 .P1 b/ equals P1 b. So P2 P1 D P1 D aaT =aT a where a D .1; 2; 0/. 33 If P1 P2 D P2 P1 then S is contained in T or T is contained in S . 34 BB T is invertible as in Problem 29. Then .AT A/.BB T / D product of r by r invertible matrices, so rank r . AB can’t have rank < r , since AT and B T cannot increase the rank. Conclusion: A (m by r of rank r ) times B (r by n of rank r ) produces AB of rank r . Problem Set 4.3, page 226 3 23 0 0     17 48 36 687 T T and b D 4 5 give A A D and A b D . 35 8 8 26 112 4 20 23 23 1 1  1 657 6 37 T T A Ab D A b gives b D x x and p D Ab D 4 5 and e D b p D 4 5 x 4 13 5 17 E D ke k2 D 44 3 2 3 23 23 10 0 1  1 61 17 C 6 8 7 This Ax D b is unsolvable 6 5 7 24 x 5 D D 4 8 5. Change b to p D P b D 4 13 5; b D 4 exactly solves 13 14 20 17 Ab D p. x 2 1 61 1 AD4 1 1 3 In Problem 2, p D A.AT A/ 1 AT b D .1; 5; 13; 17/ and e D b p D . 1; 3; 5; 3/. p e is perpendicular to both columns of A. This shortest distance ke k is 44. 4 E D .C C 0D/2 C .C C 1D 8/2 C .C C 3D 8/2 C .C C 4D 20/2 . Then @[email protected] D 2C C 2.C C D 8/ C 2.C C 3D 8/ C 2.C C 4D 20/ D 0 and @[email protected] D 1  2.C C D 8 C 3  2.C C  8/ C 4  2.C C 4D 20/ D 0. These /   3D  48 C 36 normal equations are again D . 8 26 D 112 Solutions to Exercises 47 0/2 C .C 8/2 C .C 8/2 C .C 20/2 . AT D Œ 1 1 1 1  and AT A D Œ 4 . AT b D Œ 36  and .AT A/ 1 AT b D 9 D best height C . Errors e D . 9; 1; 1; 11/. 5 E D .C 6 a D .1; 1; 1; 1/ and b D .0; 8; 8; 20/ give b D aT b=aT a D 9 and the projection is x ba D p D .9; 9; 9; 9/. Then e T a D . 9; 1; 1; 11/T .1; 1; 1; 1/ D 0 and ke k D x p 204. 7 A D Œ 0 1 3 4 T , AT A D Œ 26  and AT b D Œ 112 . Best D D 112=26 D 56=13. 8 b D 56=13, p D .56=13/.0; 1; 3; 4/. .C; D/ D .9; 56=13/ don’t match .C; D/ D .1; 4/. x Columns of A were not perpendicular so we can’t project separately to ﬁnd C and D . 2 3 23 " #" # " # 10 0"# 0 Parabola C 4 8 26 C 36 61 1 17 6 87 T D D 4 5. A Ab D 8 26 92 D D 112 . 9 Project b 4 x 1 3 95 8 4D to 3D E 26 92 338 E 400 1 4 16 20 2 32 3 2 3 23 2 3 10 0 0 C 0 C 0 Exact cubic so p D b, e D 0. 6 1 1 1 1 76 D 7 6 8 7 6 D 7 1 6 47 7 This Vandermonde matrix . 10 4 D . Then 4 5 D 4 1 3 9 27 54 E 5 4 8 5 E 28 5 gives exact interpolation 3 1 4 16 64 F 20 F 5 by a cubic at 0; 1; 3; 4 11 (a) The best line x D 1 C 4t gives the center point b D 9 when b D 2. b t (b) The ﬁrst equation C m C D T P ti D bi divided by m gives C C Db D b. tb 12 (a) a D .1; : : : ; 1/ has a a D m, a b D b1 C    C bm . Therefore b D aT b=m is the x P ba b D .1; 2; b/ ke k2 D x " # 1 111 p D .3; 3; 3/ T 111 . p e D 0. P D e D . 2 ; 1; 3/ 3 111 mean of the b ’s (c) T P 13 .AT A/ 1 A T .b 14 The matrix .b x (b) e D b Ax / D b x T x . When e D b T 1 T m i D1 .bi b/2 D variance x Ax averages to 0, so does b x T T x. 1 x /.b x / is .A A/ A .b Ax /.b Ax / A.A A/ ....
View Full Document

## This note was uploaded on 09/25/2012 for the course PHY 103 taught by Professor Minki during the Spring '12 term at Korea Advanced Institute of Science and Technology.

Ask a homework question - tutors are online