*Assignment 2

QUESTION 2. a) Solving the system using the Aconstruction we have

i)Q2ai.m ¯

A=[2 -1 3;4 2 -5;6 3 1];

b=[8;-9;12];
x=A\b

which give the output

ans =
0.50000
2.00000
3.00000

ii)Solving the system using Gauss− Seidel method. In this instance you needed to know how to
use write codes using the functions as explained in your tutorial letter. The technique to put the
Gauss− Seidel code inside the iterative− linear− solver code since it does most of the import calcu-
lations with regard to TOL. You could directly amend the Gauss− Seidel code to accommodate the
TOL. Notice also that I have slightly modified the iterative− linear− solver code. Input A, b and

xinitial= [0;0;0], TOL=10ˆ{-7}, k=200 (say)

to
iterative− linear− solve.m

function
xnew=iterative_linear_solve(A,b,xinitial,TOL,max_it)
xold=xinitial; k=0;
do xnew=Gauss_Seidel(A,b,xold); err=max(abs((xnew-xold)./xnew));
xold=xnew; k=k+1;
until((err<TOL) | (k>max_it));
k if (k>max_it) disp("ERROR: METHOD DID NOT CONVERGE");
xnew=[]; endif endfunction

and
Gauss-Seidel.m

function xnew=Gauss-Seidel(A,b,xold) n=size(A)(1);

At=A; xnew=xold; for k=1:n
At(k,k)=0;
end
for k=1:n xnew(k)=(b(k)-At(k,:)*xnew)/A(k,k);
end endfunction

we found that the Gauss− Seidel method did not converge at all and we got the answer

ERROR: METHOD DID NOT CONVERGE

2
 APM1513/202

To see this much further, take for instance, max− it=10000, the method still does not converge.
b) Solving the system using the Aconstruction we have
i) Q2bi.m ¯

A=[10 1 2;1 10 -1;2 1 10];

b=[3;1.5;-9];
x=A\b

which give the output

ans =
0.50000
0.00000
-1.00000

ii) We found that the Gauss− Seidel method did in fact converged when k = 16(number of iterations).
QUESTION 3. octave/matlab codes
Q3.m

function test=matrix(A) [n,m]=size(A) if (n==m)

disp(’matrix is a square’)
else
disp(’matrix is not a square’)
end for k=1:n
for l=1:m
if (k==l) & (abs(A(k,l))>sum(abs(A(k,1:1:m)))-abs(A(k,l)))
disp(’diagonaly dominant’)
end
if (k==l) & (abs(A(k,l))<sum(abs(A(k,1:1:m)))-abs(A(k,l)))
disp(’not diagonaly dominant’)
end
end
end end

a) We use the matrix A = [2 − 13; 42 − 5; 631] and B = [034; 133] to test if they are a square matrix
or not using the above code
i) for the matrix A the output is

n =

m =

matrix is a square

3
ii) and for matrix B output is

n =

m =

matrix is not a square

b) To test the above code for matrix diagonality we use the matrix in question 2 (b) and we indeed
see that the matrix is indeed diagonally dominant with the following output from the above code:

diagonaly dominant diagonaly

dominant diagonaly dominant

This is the results for each row computation for testing strictly diagonality condition and we have
done it for A11 , A22 , and A33 . The condition on which strictly diagonality for matrices is discussed
in your study guide page 62. The above code also works for any other matrix that is not diagonally
dominant for example take the following matrix
A = [234; 263; 986] for which the code gives the output

not diagonaly dominant diagonaly dominant not diagonaly dominant

which clearly shows that the metrix A is not diagonal as we see for the first and third rows for A11
and A33 respectively. In this case the strictly diagonality condition is not satisfied.

QUESTION 4.LEFT AS AN EXERCISE!

QUESTION 5. Q5.m

format long n=13; H=ones(n,n); b=ones(n); x=ones(n); for K=1:n

for L=1:n
H(K,L)=1/(K+L-1);

end
end

x=(Hˆn\bˆn)ˆ(1/n)

Note: the Hilbert matrix can also be generated by the single command H = hilb(8) on the Oc-
tave/Matlab command prompt. If you increase the dimension n you get the following warning:
Matrix is close to singular or badly scaled. This means that the results are not that reliable for n
large.

4
 APM1513/202

QUESTION 6. Q6.m

format long

b=[40000 44000 52000 64000 80000 84000]’;

t=[1:6]’;
A=[t.ˆ2 t ones(6,1)];
det(A’*A)

xs=A\b

tt=1:6;
s=xs(1)*t.ˆ2 + xs(2)*t + xs(3);
plot(t,b,’r’,tt,s,’b’)
err=A*xs-b

Below is the output graph is

450

400

350

300
y coordinates

250

200

150

100

50

−50
−1 0 1 2 3 4 5 6 7 8
x coordinates

Figure 1: The sales curve.

5
The smooth graph is the best quadratic fit to the data while the irregular graph is the actual once.
A similar problem can be found on your tutorial letter page. 74.
QUESTION 7. Q7.m

format
long x=[-1 0 1 2 3]’;
b=[14 -5 -4 1 22]’;
A=[x x.ˆ2 x.ˆ3] xs=A\b;
xx=-1:0.1:8; plot(x,b,’*r’,xx,xs(1)*xx+xx.ˆ2*xs(2)+xx.ˆ3*xs(3),’b’)

which give the results

450

400

350

300

250

200

150

100

50

−50
−1 0 1 2 3 4 5 6 7 8

6
 APM1513/202

QUESTION 8. 7. Here is the octave/matlab script files to do the calculations for the eigenvalues
and the eigenvectors
(a)File eig1.m

A=
[2.6731385 2.6381454 3.8272855 4.7022868;
2.7080530 -0.009461 0.1019678 2.3595453;
5.4426950 2.4321965 4.8982919 7.1372124;
2.7672696 3.4815125 2.3344020 8.2787928]
[P1 L1]=eig(A) diag=inv(P1)*A*P1

Typing the script file name > eig1 on the command prompt we obtain

A =

2.6731385 2.6381454 3.8272855 4.7022868

2.7080530 -0.009461 0.1019678 2.3595453;
5.4426950 2.4321965 4.8982919 7.1372124;
2.7672696 3.4815125 2.3344020 8.2787928

P1 =

-0.4624204 -0.4690215 -0.4356199 -0.4616522

-0.1748299 0.8048279 -0.7504984 -0.0093571
-0.6844440 0.2967166 0.3110040 -0.6972133
-0.5358529 -0.2103097 0.3876389 0.5483459
L1 =

14.78446 0 0 0
0 -2.16659 0 0
0 0 0.30142 0
0 0 0 2.92146

diag =

4.78446 0 0 0
0 -2.16659 0 0
0 0 0.30142 0
0 0 0 2.92146

You can save the above script using any name of your choice as an m file. Notice the difference
between a scrip file and a function file.
ii) Power method code is given in the tutorial letter. Just enter A, TOL and max and continue from
there.

7
iii) Matrix diagonalization is trivial in Octave, because the eigenvectors matrix P1 returned by Oc-
tave is also the diagonalization matrix as we can check by evaluating P1−1 AP1 in the script above,

(b)(a)File eig2.m

B= [2.9541 3.6650 7.1445 1.9998;

9.1406 15.3006 12.2773 10.3552;
7.9200 9.6881 7.4619 6.7253;
21.6960 35.6289 30.0172 20.7166]
[P1 L1]=eig(B) diag=inv(P1)*B*P1

Typing the script file name > eig2 on the command prompt we obtain

B =

2.9541 3.6650 7.1445 1.9998;

9.1406 15.3006 12.2773 10.3552;
7.9200 9.6881 7.4619 6.7253;
21.6960 35.6289 30.0172 20.7166]

P1 =

0.110699 0.730045 -0.708599 -0.353896

0.397694 -0.508586 0.098037 -0.461470
0.258050 0.291577 0.632058 0.230541
0.873496 -0.351222 -0.297957 0.780163
L1 =

48.55537 0.00000 0.00000 0.00000

0.00000 2.29227 0.00000 0.00000
0.00000 0.00000 -3.08485 0.00000
0.00000 0.00000 0.00000 -1.32959
diag =

48.55537 0.00000 0.00000 0.00000

0.00000 2.29227 0.00000 0.00000
0.00000 0.00000 -3.08485 0.00000
0.00000 0.00000 0.00000 -1.32959

You can save the above script using any name of your choice as an m file. Notice the difference
between a scrip file and a function file.
ii) Power method code is given in the tutorial letter. Just enter A, TOL and max and continue from
there.
iii) Matrix diagonalization is trivial in Octave, because the eigenvectors matrix P1 returned by Oc-
tave is also the diagonalization matrix as we can check by evaluating P1−1 AP1 in the script above,

8
 APM1513/202

QUESTION 9.
Here is the modified Octave script file
power− method.m

function [e_vec lam]=power_method(A,TOL,max_it) k=0; n=size(A)(1);

A_inv=inv(A);% modified here e_vec_old=rand(n,1);
do e_vec_new=A_inv*e_vec_old;
lam=(e_vec_new’*e_vec_old)/(e_vec_old’*e_vec_old);
err=norm(e_vec_new/norm(e_vec_new)-e_vec_old/norm(e_vec_old));
e_vec_old=e_vec_new;
k=k+1; until((err<TOL) | (k>max_it));
k e_vec=e_vec_new/norm(e_vec_new);
if (k>max_it) disp("ERROR: METHOD DID NOT CONVERGE");
e_vec=[]; lam=[]; endif endfunction

which give the following result

k=36 answer = 0.33752

0.45936
-0.25171
-0.78212

for the following matrix

B=[-1.54575 -3.47002 -1.70112 -2.58917;

-3.28104 -2.07998 -1.45597 -2.75629;
0.55497 0.94078 2.02863 0.46100;
8.94120 9.67047 4.47796 9.09710]

This is indeed the smallest eigenvector as it can be confirmed from the above results obtained by
eig2.m code. For the following matrix

B=[4.9541 6.6650 8.1445 2.9998; 10.1406 17.3006 14.2773 9.3552; 8.9200

10.6881 8.4619 7.7253; -25.6960 -40.6289 -38.0172 -21.71666],

we obtain

k=19 answer = 0.564503

-0.738274
0.365124
0.054515

which give a contradictory results, i.e. in question 3. we have eig3.m code giving complex results
while here we have real results. Because of this fact, we cannot say for sure if this results are for
the lowest eigenvector. This has something to do with the fact that the power− method did not work
in question 3. That also means here that the power− method failed to working.