This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: MA 35100 HOMEWORK ASSIGNMENT #12 SOLUTIONS Problem 1. pg. 229; prob. 8 Consider a linear transformation L ( ~x ) = A~x from R n to R m , with ker( L ) = { ~ } . The pseudoin verse L + of L is the transformation from R m to R n given by L + ( ~ y ) = ( the leastsquares solution of L ( ~x ) = ~ y ) . a. Show that the transformation L + is linear. Find the matrix A + of L + , in terms of the matrix A of L . b. If L is invertible, what is the relationship between L + ad L 1 ? c. What is L + ( L ( ~x ) ) , for ~x in R n ? d. What is L ( L + ( ~ y ) ) , for ~ y in R m ? e. Find L + for the linear transformation L ( ~x ) = 1 0 0 1 0 0 ~x. Solution: (a) Consider the linear system A~x = ~ y . According to Theorem 5.4.6 on page 224 of the text, the leastsquares solution is ~x * = ( A T A ) 1 A T ~ y . That means we have the transformation L + ( ~ y ) = ~x * = A + ~ y in terms of A + = ( A T A ) 1 A T . Using Definition 2.1.1 on page 44, we see that L + is indeed a linear transformation. (b) Assume now that L is invertible, so that the matrix A is invertible. Theorem 5.3.9 on page 216 states rank( A T ) = rank( A ), and Theorem 3.1.7 on page 109 states that an n n matrix is invertible if and only if its rank is n , so A T is also invertible. We may now use Theorem 2.4.7 on page 83: A + = ( A T A ) 1 A T = A 1 ( A T ) 1 A T = A 1 . Hence A + = A 1 , so that L + = L 1 if L is invertible. (c) For all ~x R n we have L + ( L ( ~x ) ) = ( A T A ) 1 A T ( A~x ) = ( A T A ) 1 ( A T A ) ~x = ~x. Hence L + ( L ( ~x ) ) = ~x . (d) For all ~ y R m we have L ( L + ( ~ y ) ) = A ( A T A ) 1 A T ~ y = A ( A T A ) 1 A T ~ y. We consider Theorem 5.4.7 on page 225. Write A = ~v 1 ~v 2 ~v n ; its column vectors are linearly independent by Theorem 3.2.8 on page 119 because ker( A ) = { ~ } . According to Theorem 3.1.3 on page 105 they form a basis for the subspace V = im( A ) = span( ~v 1 ,~v 2 ,...,~v n ) of R m . Hence we conclude from Theorem 5.4.7 that A ( A T A ) 1 A T is the matrix of the orthogonal projection onto V . This means L ( L + ( ~ y ) ) = proj V ~ y is the projection of ~ y onto V = im( A ). 1 (e) The matrix A + of L + is A + = ( A T A ) 1 A T = 1 0 0 0 1 0 1 0 0 1 0 0  1 1 0 0 0 1 0 = 1 0 0 1 1 1 0 0 0 1 0 = A T . Hence the pseudoinverse of L is L + ( ~ y ) = 1 0 0 0 1 0 ~ y . Problem 2. pg. 230; prob. 15 Consider an m n matrix A with ker( A ) = { ~ } . Show that there exists an n m matrix B such that B A = I n . Hint : A T A is invertible. Solution: If A is an m n matrix then A T is an n m matrix, so that A T A is an n n square matrix. According to Theorem 5.4.2 on page 221 of the text, since ker( A ) = { ~ } we may conclude that A T A is invertible. Consider the n m matrix B = ( A T A ) 1 A T , as motivated by Exercise 5.4.8 on page 229 (i.e., Problem 1 above). Then we have B A = ( A T A ) 1 ( A T A ) = I n ....
View
Full
Document
This note was uploaded on 03/31/2012 for the course MA 351 taught by Professor ?? during the Spring '08 term at Purdue UniversityWest Lafayette.
 Spring '08
 ??
 Linear Algebra, Algebra

Click to edit the document details