20095ee103_1_hw6_sols

20095ee103_1_hw6_sols - L. Vandenberghe 11/12/09 EE103...

Info iconThis preview shows pages 1–4. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: L. Vandenberghe 11/12/09 EE103 Homework 6 solutions 1. Exercise 9.9. The first plot is for = 1, the second plot is for = 100, the last plot is for = 10000. For = 1, there is not enough smoothing, for = 10000 we have too much smoothing, and = 100 seems about right. 200 400 600 800 1000-0.5 0.5 200 400 600 800 1000-0.5 0.5 200 400 600 800 1000-0.5 0.5 i i x x x 2. Exercise 9.11. (a) We use the property that a square matrix B is nonsingular if and only if Bx = 0 implies x = 0. Suppose x R m and y R n satisfy bracketleftBigg I A A T bracketrightBigg bracketleftBigg x y bracketrightBigg = 0 . In other words, x + Ay = 0 and A T x = 0. Then A T x = A T Ay = 0 . A T A is positive definite (hence, nonsingular) if A has a zero nullspace. Therefore y = 0, and from x = Ay we can also conclude that x = 0. 1 (b) The equations bracketleftBigg I A A T bracketrightBigg bracketleftBigg x y bracketrightBigg = bracketleftBigg b bracketrightBigg are equivalent to x + Ay = b, A T x = 0 . We have to check that those two equations are satisfied for x = b Ax ls , y = x ls . Clearly this choice of x and y satisfies x + Ay = b . Moreover, we have A T x = A T ( b A ( A T A )- 1 A T b ) = A T b ( A T A )( A T A )- 1 A T b = A T b A T b = 0 . 3. Exercise 10.4. (a) The second problem is a least-squares problem minimize vextenddouble vextenddouble vextenddouble vextenddouble vextenddouble bracketleftBigg A c T bracketrightBigg y bracketleftBigg b d bracketrightBiggvextenddouble vextenddouble vextenddouble vextenddouble vextenddouble . We need to verify that y satisfies the normal equations ( A T A + cc T ) y = A T b + dc. We know that A T A x = A T b , so ( A T A + cc T ) x = A T b + ( c T x ) c. Also, ( A T A + cc T )( A T A )- 1 c = c + ( c T ( A T A )- 1 c ) c. Therefore ( A T A + cc T ) y = ( A T A + cc T ) parenleftBigg x + d c T x 1 + c T ( A T A )- 1 c ( A T A )- 1 c parenrightBigg = A T b + ( c T x ) c + d c T x 1 + c T ( A T A )- 1 c (1 + c T ( A T A )- 1 c ) c = A T b + ( c T x ) c + ( d c T x ) c = A T b + dc. 2 (b) If we plug in the QR factorization of A , the formulas for x and y reduce to x = R- 1 Q T b y = x + d c T x 1 + c T R- 1 R- T c R- 1 R- T c. QR factorization of A (2 mn 2 flops). Compute u = Q T b (2 mn flops). Compute x = R- 1 u using backward substitution ( n 2 flops). Compute w = R- 1 R- T c using forward and backward substitution (2 n 2 flops). Compute y = x + d- c T x 1+ c T w w (6 n flops). The total is 2 mn 2 + 2 mn + 3 n 2 ....
View Full Document

Page1 / 9

20095ee103_1_hw6_sols - L. Vandenberghe 11/12/09 EE103...

This preview shows document pages 1 - 4. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online