You are on page 1of 10
‘Numerical Linear Algebra Name (Print): Key __ _ all 2017 Exam 2 11/03/17 ‘This exam contains 9 pages (including this cover pags) and 8 problems. Check to se if any’ pages are ulesing. Enter all requested information on the top of this page top of every pga it ease the pages becouue separated, ‘The following rues appl If you use a “fundamental theorem” you ‘muistndieate this aud explain why the theorem may be applied ‘+ Organize your work, in a reasonably neat and coherent way, in the space provided. Work seat~ tered all over the page without a clear ordering ‘will ocsive very litle credit, + Short answer questions: Questions Inbeled se “Short Answer” can be answered by skmply writ- ing an oquation or m sentence ot appropriately drawing a figure. No calculations are necessary or texpectad for these problems, + Unless the question is specified as short an- wer, mysterious or unsupported answers right not receive full eredit. An incorrect nssver supported by substantially corret eal Jations and explanations might sil recive partial credit, Do not write inthe table tothe right. ane put your iitials on the | Problem | Points | Score 2 | a fe «fo s [as : 6 15 7 [6 = | o ‘tat: | 100 Numeral Linear Algebra Bram 2- Page 2 of yay 1, (10 points) Suppote (4: R-# Rsatisy ste) =0, ioe) =0, iy £0) = 00 ‘in, g(2) =00, Be) on Je) PB ae) (2) (8 points) 1s 462) = O(o(2)) an x > 0? Is ols) = O(f(2)) as 207 Briey explain Example’ Example SO=OGO) a. x>0. FOR OH) (b) (5 points) Is f(x) = O(9(r)) a3 7 + 20? Is g(x) = O(f(+)) as x + 00? Briefly explain. Example ' SOx)=xt SOO OGi) as x20. Jiniek gor# SA) erical Linear Algebra Bram 2 -Page 3 suf) 1 2. (15 points) Cousider an algorithm for the problem of computing the (fll) QR factorization of fa matrix A RM". The data for this problem is a matrix A, and the solution is a unitary tmatrix Q and an upper triangular matrix R, Can this algorithm be backward stable? Explain. Ths cannot be backwerds steble since & will have Poondo£& error and thes nor be Unitary, ‘Numerical Linear Algebra Bxam 2 Pago dof 0 11/os/37 5 (15 points) Consider the following Matlab code provided below. Find an asymptotic formula for the nutaber of flops for this algorithm. Fonction [HC] = Bard B44 Hato t00p fhe Fortney 5 oh = Ringers > nome l yh additions wna EAS mol Sut ractioes Beep Srev/noem Slops~ 3 Shy faa “Dy j= ~ dn Nema Liat Als Brain 9- Page & of yan 4. (8 pos) Consider the following problen: sey=2ee1 sd the flloring slept fr computing ths problns on 8 canpult Fle) = 2 @n(z)) 21. 1s hs grin buds tae? 1 hs ain stab? So )al2 OF100)O) = (2 @xtlt£))@! = (2xe4,)U§)) 1 =Bxcite, U4) +i] +8) = 2xC\FE LE) UH)+ & +1 Ke XCOUFAMIFE CHE 8, => TW =50h). * wevey, WS XU) FEL TY WxIT “ Which diverges as Ix\l~o. Lee K= xc fui Wit h)e Then, WE 03-S te) <1 etl: = OF.) NSCON jfaxar ENTE YH) +5411 Ase, Next = OE). Therefore, the algerithm is Farword Stable, ‘Numerical Linear Algebra Exam 2- Page 6 of 9 uijosyt 5. (15 points) Suppose A € B® has the following singular value decomposition: noo 0 Aculoa ofv 0 0 0 ad Z € Bis given by solute error wth by the algorithm IAF is computed using respect tothe two norm KIAV= 10 o> = I Sb0- Foul) < USGant Now, ll Sool= Axis 10 = IlSa-Fbolle [ol Is stable alzorit in, lat the cet vale A andthe vector co Numerical Linear Algebra Exam 2- Page 7 of 9 1yos/it 6. (15 polats) Given the vector #'¢ R} defined by “lh construct a unitary matrix Q, ie, a Householder relector, such that % “1 4 Numerical Linear Algebra Ea Page 8 of nya 7. (15 points) For students who are not MST graduate students (0) (0 pis) Suppose A Bi nonsingular. Prove tat the rdatie condition ber 8(A) = JANA™?|] relative to the two norm satisfies: (A) 2 1. K(A)= & > 7% =), oF (8) (pit) Fora vr simple reson the ative sodton miner for ny probleneaot bbe less than 1. Explain why this mist be true by assuanng there exits backwards stable algorithms forthe problem TS KOM) less then 1 this would imply Vhet a backwards stable algorithn Veld pe Inore accurate than F, Ths ts tmporsible: Tn Pace Pe definition ot KOS Constrvcted precisely 50 that Hhis 5 Peposrible: ‘Numerical Li soto nist Algebra Bam 8, (15 polnts) For MST graduate students only. Othe for potential bonus points. tudents can do this problem (a) (10 points) Suppose A R" fe nonsingular and p € R satisfies 1 < p < co. Provethat the condition maaber x(4) = [A[47H] relative to the norm =p satis nA) 21 4S NIN=NA AMS WAIMAY =A) (&) (8 points) Fora vary simple reason the relative condition number for any problem cannot ‘bes than 1. Explain why this must be true by assunug there exists a backwatds stable algorithm forthe problem, See Preble 47-

You might also like