Introduction to Linear algebra-Strang-Solutions-Manual_ver13

# 0 12 1 c 1 4 1 12 2 c 2 1 2 1 d 1 4 2 1 2 5 mean

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: c D .2; 3; 8/ the Ph.D. takes 4 hours .x1 C x2 C 2x3 D 4/ and charges \$8. The teacher in the dual problem now has y  2; y  3; 2y  8 as constraints AT y  c on the charge of y per problem. So the dual has maximum at y D 2. The dual cost is also \$8 for 4 problems and maximum D minimum. 7 x D .2; 2; 0/ is a corner of the feasible set with x1 C x2 C 2x3 D 4 and the new constraint 2x1 C x2 C x3 D 6. The cost of this corner is c T x D .5; 3; 8/  .2; 2; 0/ D 16. Is this the minimum cost? Compute the reduced cost r if x3 D 1 enters (x3 was previously zero). The two constraint equations now require x1 D 3 and x2 D 1. With x D .3; 1; 1/ the new Solutions to Exercises 86 cost is 3:5 optimal. 1:3 C 1:8 D 20. This is higher than 16, so the original x D .2; 2; 0/ was Note that x3 D 1 led to x2 D 1 and a negative x2 is not allowed. If x3 reduced the cost (it didn’t) we would not have used as much as x3 D 1. 8 y T b  y T Ax D .AT y /T x  c T x . The ﬁrst inequality needed y  0 and Ax b  0. Problem Set 8.5, page 451 1 R 2 0 h i2 sin..j Ck/x/ j Ck 0 R 2 D 0 and similarly 0 cos..j k /x/ dx D 0 R 2 k ¤ 0 in the denominator. If j D k then 0 cos2 jx dx D  . cos..j C k/x/ dx D Notice j 2 Three integral tests show that 1; x; x 2 1 3 are orthogonal on the interval Œ 1; 1 : R1 1 .1/.x/ dx D 0; 1 .1/.x dx D 0; 1 .x/.x 2 / dx D 0: Then 1 3 1 2 2 2 2 2x D 2.x / C 0.x/ C 3 .1/. Those coefﬁcients 2; 0; 3 can come from integrating 3 f .x/ D 2x 2 times the 3 basis functions and dividing by their lengths squared—in other words using aT b=aT a for functions (where b is f .x/ and a is 1 or x or x 2 1 ) exactly 3 as for vectors. p 1 3 One example orthogonal to v D .1; 2 ; : : :/ is w D .2; 1; 0; 0; : : :/ with kwk D 5. R1 R1 1 3 4 c x/ dx D 0 and 1 .x 2 3 /.x 3 c x/ dx D 0 for all c (odd functions). 1 .1/.x R1 Choose c so that 1 x.x 3 c x/ dx D Œ 1 x 5 c x 3 1 1 D 2 c 2 D 0. Then c D 3 . 5 3 5 3 5 R1 R1 2 1 / 3 5 The integrals lead to the Fourier coefﬁcients a1 D 0, b1 D 4= , b2 D 0. 6 From eqn. (3) ak D 0 and bk D 4=k (odd k ). The square wave has kf k2 D 2 . 1 Then eqn. (6) is 2 D .16= 2 /. 12 C 31 C 51 C    /. That inﬁnite series equals  2 =8. 2 2 1; 1 odd square wave is f .x/ D x=jx j for 0 &lt; jx j &lt;  . Its Fourier series in equation (8) is 4= times Œsin x C .sin 3x/=3 C .sin 5x=5/ C   . The sum of the ﬁrst N terms has an interesting shape, close to the square wave except where the wave jumps between 1 and 1. At those jumps, the Fourier sum spikes the wrong way to ˙1:09 (the Gibbs phenomenon) before it takes the jump with the true f .x/: 7 The This happens for the Fourier sums of all functions with jumps. It makes shock waves hard to compute. You can see it clearly in a graph of the sum of 10 terms. p 8 kvk2 D 1 C 1 C 1 C 1 C   D 2 so kvk D 2; kvk2 D 1 C a2 C a4 C   D 1=.1 a2 / 2 4 8R p p 2 so kvk D 1= 1 a2 ; 0 .1 C 2 sin x C sin2 x/ dx D 2 C 0 C  so kf k D 3 . 9 (a) f .x/ D .1 C square wave/=2 so the a’s are 2 =3 , 0, 2=5 , . . . (...
View Full Document

Ask a homework question - tutors are online