polynomial interpolation a x

Size: px

Start display at page:

Download "polynomial interpolation a x"

Augustine Marsh
5 years ago
Views:

3 8 polynomal nterpolaton a x

4 5 polynomal nterpolaton b x

5 output.txt Oct 17, 11 5:21 Page 1/1 >> prob1 Part a z0 z1 z2 teratons root the remanng root s conjugate of the prevous one double check wth Matlab: Part b z0 z1 z2 teratons root the remanng root s conjugate of the prevous one double check wth Matlab: >> prob3 B&F a and a degree Lagrange Newton err bound e e e e e e e e e 03 B&F b and b degree Lagrange Newton err bound e e e e e e e e e 02 >> Oct 17, 11 5:21 Page 1/1 HW 4 Problem 1 B&F a,b tol = 1e 5; maxt = 20; format long fprntf( Part a \n ); p = [ ]; fprntf( 13s 13s 13s 8s 13s\n, z0, z1, z2, teratons, root ); z0=1; z1=2; z2=3; fprntf( 13.8f 13.8f 13.8f 8d 13.8f f\n,z0,z1,z2,k,real(z),mag(z)); z0= 4; z1= 4.5; z2= 5; fprntf( 13.8f 13.8f 13.8f 8d 13.8f f\n,z0,z1,z2,k,real(z),mag(z)); z0=0; z1=1; z2=2; fprntf( 13.8f 13.8f 13.8f 8d 13.8f f\n,z0,z1,z2,k,real(z),mag(z)); fprntf( the remanng root s conjugate of the prevous one\n ); fprntf( double check wth Matlab:\n ); dsp(roots(p)); prob1.m fprntf( Part b \n ); p = [ ]; fprntf( 13s 13s 13s 8s 13s\n, z0, z1, z2, teratons, root ); Prnted by Fernando Guevara Vasquez z0= 5; z1= 4; z2= 3; fprntf( 13.8f 13.8f 13.8f 8d 13.8f f\n,z0,z1,z2,k,real(z),mag(z)); z0=3; z1=4; z2=5; fprntf( 13.8f 13.8f 13.8f 8d 13.8f f\n,z0,z1,z2,k,real(z),mag(z)); z0=0; z1=1; z2=2; fprntf( 13.8f 13.8f 13.8f 8d 13.8f f\n,z0,z1,z2,k,real(z),mag(z)); fprntf( the remanng root s conjugate of the prevous one\n ); fprntf( double check wth Matlab:\n ); dsp(roots(p)); Monday October 17, 2011 output.txt, prob1.m 1/5

6 Oct 16, 11 1:49 Page 1/3 HW4 Problem 3 B&F a and a fprntf( B&F a and a \n ); nterpolaton node abscssas (we reordered st the node at whch we evaluate the nterpolaton s always nsde the nterval) xd = [0.25,0.5,0.75,0]; ordnates yd = [ , , ,1]; evaluaton pont x = 0.43; true functon (USING VECTORIZED OPERATIONS) f exp(2*x); The functon and frst few dervatves are: f(z) = exp(2*x) f (z) = 2*exp(2*x) f (z) = 4*exp(2*x) f (z) = 8*exp(2*x) f (z) = 16*exp(2*x) here we gve the part of the bound dependng on f. It would be better to fnd the max of ths functons over the nterpolaton nterval [0,0.75] for nterp poly of degree 1 we bound f fmax(1) = 4*exp(2*0.75); for nterp poly of degree 2 we bound f fmax(2) = 8*exp(2*0.75); for nterp poly of degree 2 we bound f fmax(3) = 16*exp(2*0.75); vectors to store nterpolaton polynomal evaluated at x and error lagr = zeros(3,1); newt = zeros(3,1); err = zeros(3,1); for k=1:3, loop on degree Lagrange nterpolaton lttle trck to take only the frst k nodes lagr(k) = lagrange(xd(1:k+1),yd(1:k+1),x); Newton nterpolaton usng dvded dfferences d = newtondd(xd(1:k+1),yd(1:k+1)); newt(k) = newtonev(xd(1:k+1),d,x); compute error bound err(k) = fmax(k)/factoral(k+1); for j=1:k+1; err(k) = err(k)*(x xd(j)); for j for k prob3.m dsplay errors fprntf( 10s 13s 13s 13s\n, degree, Lagrange, Newton, err bound ); for k=1:3, fprntf( 10d 13.4e 13.4e 13.4e\n, k,abs(lagr(k) f(x)),abs(newt(k) f(x)),abs(err(k))); for k Show all three nterpolaton polynomals n a plot fgure(1); IF DOING MORE THAN ONE FIGURE, CHANGE FIGURE NUMBER clf; clears fgure xs = lnspace(0,1); equally space ponts n an nterval contanng nterp. po Oct 16, 11 1:49 Page 2/3 nts we are gong to do several plots, so we tell Matlab to accumulate all the plot commands hold on; Plots the true functon plot(xs,f(xs), r ); Plots the nterpolaton nodes plot(xd,yd, * ); Plot the nterpolaton polynomals for k=1:3, Newton nterpolaton usng dvded dfferences d = newtondd(xd(1:k+1),yd(1:k+1)); newt = newtonev(xd(1:k+1),d,xs); plot(xs,newt, b ); We don t want to add anymore plots to the current fgure hold off; put some labels xlabel( x ); ttle( polynomal nterpolaton a ); prnt to a fle prnt( depsc2, prob3_a.eps ); B&F b and b fprntf( B&F b and b \n ); prob3.m nterpolaton node abscssas (we reordered st the node at whch we evaluate the nterpolaton s always nsde the nterval) xd = [ 0.25,0.25, 0.5,0.5]; ordnates yd = [ , , , ]; evaluaton pont x = 0; true functon (USING VECTORIZED OPERATIONS) f x.^4 x.^3 + x.^2 x +1; The functon and frst few dervatves are: f(z) = z^4 z^3 + z^2 z +1 f (z) = 4*z^3 3*z^2 + 2*z 1 f (z) = 12*z^2 6*z + 2 f (z) = 24*z 6 f (z) = 24 here we gve the part of the bound dependng on f. It would be better to fnd the max of ths functons over the nterpolaton nterval [0,0.75] for nterp poly of degree 1 we bound f fmax(1) = 12*( 0.5)^2 6*( 0.5) + 2; for nterp poly of degree 2 we bound f fmax(2) = 24*0.5 6; for nterp poly of degree 2 we bound f fmax(3) = 24; vectors to store nterpolaton polynomal evaluated at x and error lagr = zeros(3,1); newt = zeros(3,1); err = zeros(3,1); for k=1:3, loop on degree Lagrange nterpolaton lttle trck to take only the frst k nodes lagr(k) = lagrange(xd(1:k+1),yd(1:k+1),x); Prnted by Fernando Guevara Vasquez Monday October 17, 2011 prob3.m 2/5

7 Oct 16, 11 1:49 prob3.m Page 3/3 Newton nterpolaton usng dvded dfferences d = newtondd(xd(1:k+1),yd(1:k+1)); newt(k) = newtonev(xd(1:k+1),d,x); compute error bound err(k) = fmax(k)/factoral(k+1); for j=1:k+1; err(k) = err(k)*(x xd(j)); for j for k dsplay errors fprntf( 10s 13s 13s 13s\n, degree, Lagrange, Newton, err bound ); for k=1:3, fprntf( 10d 13.4e 13.4e 13.4e\n, k,abs(lagr(k) f(x)),abs(newt(k) f(x)),abs(err(k))); for k Show all three nterpolaton polynomals n a plot fgure(2); clf; clears fgure xs = lnspace( 1,1); equally space ponts n an nterval contanng nterp. p onts we are gong to do several plots, so we tell Matlab to accumulate all the plot commands hold on; Plots the true functon plot(xs,f(xs), r ); Plots the nterpolaton nodes plot(xd,yd, * ); Plot the nterpolaton polynomals for k=1:3, Newton nterpolaton usng dvded dfferences d = newtondd(xd(1:k+1),yd(1:k+1)); newt = newtonev(xd(1:k+1),d,xs); plot(xs,newt, b ); We don t want to add anymore plots to the current fgure hold off; put some labels xlabel( x ); ttle( polynomal nterpolaton b ); prnt to a fle prnt( depsc2, prob3_b.eps ); Oct 17, 11 5:19 Page 1/1 functon [z3,k] = muller(z0,z1,z2,p,maxt,tol) Fnds a polynomal root gven the ntal guesses z0,z1,z2 usng a quadratc mode (Muller s method). Ths method allows convergence to a complex root even f the z0,z1,z2 are real. Inputs z0,z1,z2 ntal guesses p vector of polynomal coeffcents, from hghest to lowest degree maxt maxmum number of teraton tol tolerance Outputs z3 approxmaton of the root k number of teratons functon [z3,k] = muller(z0,z1,z2,p,maxt,tol) [q,fz0] = horner(p,z0); [q,fz1] = horner(p,z1); [q,fz2] = horner(p,z2); for k=1:maxt, compute coeffcents of quadratc c = fz2; h1 = z1 z0; h2 = z2 z1; d1 = (fz1 fz0)/h1; d2 = (fz2 fz1)/h2; a = (d2 d1)/(z2 z0); b = d2 +h2*a; double check that a,b,c are OK qd a*(z z2)^2 + b*(z z2) + c; quadratc model [qd(z2) fz2, qd(z1) fz1, qd(z0) fz0] quadratc model should nterp f delta = sqrt(b^2 4*a*c); choose root wth smallest magntude f (abs(b delta)<abs(b+delta)) z3 = z2 + ( b+delta)/2/a; else z3 = z2 + ( b delta)/2/a; check convergence f abs(z3 z2)<tol break; prepare next teraton [q,fz3] = horner(p,z3); muller.m f (abs(fz3)<tol) break; fz0=fz1; z0=z1; fz1=fz2; z1=z2; fz2=fz3; z2=z3; Prnted by Fernando Guevara Vasquez Monday October 17, 2011 prob3.m, muller.m 3/5

8 Prnted by Fernando Guevara Vasquez Sep 28, 11 9:15 horner.m Page 1/1 Horner s algorthm to dvde a polynomal p by the lnear factor z z0.e. p(z) = q(z) (z z0) + r Inputs p vector of polynomal coeffcents, from hghest to lowest degree z0 root of lnear factor Outputs q vector of polynomal coeffcents for the quotent polynomal r resdual functon [q,r] = horner(p,z0) n = length(p); q = zeros(sze(p)); q(1) = p(1); for k=1:n 1, q(k+1) = p(k+1) + z0 * q(k); r = q(n); q = q(1:n 1); Sep 28, 11 9:14 Page 1/1 functon y = lagrange(xd,yd,x) Evaluates Lagrange nterpolaton polynomal at some ponts Inputs: xd absssa of data ponts yd ordnate of data ponts x abscssa where we want to evaluate Lagrange nterp. polynomal Outputs: y p(x) One way ot test ths s to run the followng: x = lnspace(0,5); xd = [ ]; yd = [ ]; y = lagrange(xd,yd,x) ; plot(x,y,xd,yd, r+ ) functon y = lagrange(xd,yd,x) y = zeros(sze(x)); for =1:length(xd), yb = yd(); for j=1:length(xd), f (j ) yb = yb.* (x xd(j))/(xd() xd(j)); y = y + yb; lagrange.m Monday October 17, 2011 horner.m, lagrange.m 4/5

9 Prnted by Fernando Guevara Vasquez newtondd.m Sep 28, 11 9:14 Page 1/1 computes Newton s dvded dfferences, needs only a vector for storage Inputs x nterpolaton nodes f functon values Outputs d coeffcents of Newton nterpolaton polynomal One way ot test ths s to run the followng: x = lnspace(0,5); xd = [ ]; yd = [ ]; d = newtondd(xd,yd); y = newtonev(xd,d,x); plot(x,y,xd,yd, r+ ) functon d = newtondd(x,f) n = length(x) 1; f (length(f) n+1) fprntf( length of nput vectors must be the same ); d = f; for j=1:n, for =n: 1:j, d(+1) = (d(+1) d())/(x(+1) x( j+1)); Sep 28, 11 9:14 Page 1/1 functon y = newtonev(xd,d,x) uses Horner s algorthm to evaluate the Newton nterpolaton polynomal at the ponts x, where the nterpolaton nodes are gven n xd Inputs xd nterpolaton nodes d coeffcents of the Newton nterpolaton polynomal x ponts of evaluaton of the Newton nterpolaton polynomal Outputs y p(x) functon y = newtonev(xd,d,x) santy check f (length(d) length(xd)) error( there must be as many dvded dfferences as nterpolaton nodes ); n = length(d); y = zeros(sze(x)); Horner s algorthm y = d(n); for k=n 1: 1:1, y = (x xd(k)).*y + d(k); newtonev.m Monday October 17, 2011 newtondd.m, newtonev.m 5/5

Review of Taylor Series. Read Section 1.2

Review of Taylor Series. Read Section 1.2 Revew of Taylor Seres Read Secton 1.2 1 Power Seres A power seres about c s an nfnte seres of the form k = 0 k a ( x c) = a + a ( x c) + a ( x c) + a ( x c) k 2 3 0 1 2 3 + In many cases, c = 0, and the