ASSINGMENT – 5 AND 6

NILOY KUMAR DUTTA 421MA2176

Q(1).
%Gauss Jacobbi method for solving system of linear equations %
A=[4 1 -1 1;1 4 -1 -1;-1 -1 5 1;1 -1 1 3];
B=[-2;-1;0;1];
P=input('Enter the initial roots = ');
n=50;
e=0.001;
N=length(B);
X=zeros(N,1);
for j=1:n
for i=1:N
X(i)=(B(i)/A(i,i))-(A(i,[1:i-1,i+1:N])*P([1:i-1,i+1:N]))/A(i,i);
end
if abs(X-P)<e
break
end
P=X;
end
fprintf('Iteration no = %d\n',j);
fprintf('The solution of the system of equation is =\n ');
disp(X);

Output-

Enter the initial roots = [0;0;0;0]

Iteration no = 12

The solution of the system of equation is =

-0.7521

0.0403

-0.2803

0.6901

%Gauss Jacobbi method for solving system of linear equations %

A=[4 1 -1 1;1 4 -1 -1;-1 -1 5 1;1 -1 1 3];
B=[-2;-1;0;1];
P=input('Enter the initial roots = ');
n=input('Enter the number of iterations = ');
N=length(B);
X=zeros(N,1);
for j=1:n
for i=1:N
X(i)=(B(i)/A(i,i))-(A(i,[1:i-1,i+1:N])*P([1:i-1,i+1:N]))/A(i,i);
end
 k=abs(X-P);
P=X;
end
fprintf('The solution of the system of equation is =\n ');
disp(X);
fprintf('The error is =\n');
disp(k);

Output-

Enter the initial roots = [0;0;0;0]

Enter the number of iterations = 10

The solution of the system of equation is =

-0.7497

0.0389

-0.2794

0.6871

The error is =

0.0019

0.0027

0.0021

0.0019

Enter the initial roots = [0;0;0;0]

Enter the number of iterations = 20

The solution of the system of equation is =

-0.7534

0.0411

-0.2808

0.6918

The error is =

1.0e-04 *
 0.1443

0.1333

0.0971

0.1684

Table-

After 10 Iterations After 20 Iterations

0.0019 0.00001443

Error 0.0027 0.00001333

0.0021 0.00000971

0.0019 0.00001684

Observation-

If we increase the no of iterations then we get more accurate solution of a system of equation

By using Gauss Jacobi method.

Q(2).
%Gauss Seidel method for solving system of linear equations %
A=[4 1 -1 1;1 4 -1 -1;-1 -1 5 1;1 -1 1 3];
B=[-2;-1;0;1];
P=input('Enter the initial roots = ');
n=50;
e=0.001;
N=length(B);
X=zeros(N,1);
Y=zeros(N,1);
for j=1:n
for i=1:N
X(i)=(B(i)/A(i,i))-(A(i,[1:i-1,i+1:N])*P([1:i-1,i+1:N]))/A(i,i);
P(i)=X(i);
end
if abs(Y-X)<e
break
end
Y=X;
end
fprintf('Iteration no = %d\n',j);
fprintf('The solution of the system of equation is =\n ');
disp(X);

Output-

Enter the initial roots = [0;0;0;0]

Iteration no = 8

The solution of the system of equation is =

-0.7532

0.0410

-0.2807

0.6916

%Gauss Seidel method for solving system of linear equations %

A=[4 1 -1 1;1 4 -1 -1;-1 -1 5 1;1 -1 1 3];
B=[-2;-1;0;1];
P=input('Enter the initial roots = ');
n=input('Enter the number of iterations = ');
N=length(B);
X=zeros(N,1);
Y=zeros(N,1);
for j=1:n
for i=1:N
X(i)=(B(i)/A(i,i))-(A(i,[1:i-1,i+1:N])*P([1:i-1,i+1:N]))/A(i,i);
P(i)=X(i);
end
k=abs(Y-X);
Y=X;
end
fprintf('The solution of the system of equation is =\n ');
disp(Y);
fprintf('The error is =\n');
disp(k);

Output-

Enter the initial roots = [0;0;0;0]

Enter the number of iterations = 10

The solution of the system of equation is =

-0.7534
 0.0411

-0.2808

0.6918

The error is =

1.0e-04 *

0.5689

0.1956

0.2685

0.3444

Enter the initial roots = [0;0;0;0]

Enter the number of iterations = 20

The solution of the system of equation is =

-0.7534

0.0411

-0.2808

0.6918

The error is =

1.0e-08 *

0.1822

0.0626

0.0860

0.1103

Table-

After 10th Iterations After 20th Iterations

 0.00005689 0.1822e-08

0.00001956 0.0626e-08

Error 0.00002685 0.0860e-08

0.00003444 0.1103e-08

Observation-

Here error is inversely proportional to the no. of Iterations, Hence if we increase the no of
iterations then we get more accurate solution of a system of equation by using Gauss seidel method.

Comparison of convergence speed-

Gauss Seidel method (took 8 iterations to get accuracy) converges faster than Gauss Jacobi
method(took 12 iterations to get accuracy).

Q(3).

(a) Trapezoidal Rule-

f=@(x)(1/(1+(x.^2)));
a=0;
b=1;
n=input('The number of intervals =');
h=(b-a)/n;
m=0;
for i=1:n-1
m=m+f(a+(i*h));
end
s=(h/2)*(f(a)+f(b)+2*m);
k=4*s;
fprintf('The required value of pi using Trapezoidal rule = %4.8f',k);
e=abs(pi-k);
fprintf('\n The error is = %4.8f',e);

Output-

The number of intervals =6

The required value of pi using Trapezoidal rule = 3.13696307

The error is = 0.00462959gut

The number of intervals =12

The required value of pi using Trapezoidal rule = 3.14043525

The error is = 0.00115741

The number of intervals =24

The required value of pi using Trapezoidal rule = 3.14130330

The error is = 0.00028935

(b) Simpson 1/3 Rule-

f=@(x)(1/(1+(x.^2)));
a=0;
b=1;
n=input('The number of intervals=');
h=(b-a)/n;
m1=0;
m2=0;
for i=1:2:n-1
t=a+(i*h);
m1=m1+f(t);
end
for i=2:2:n-2
t=a+(i*h);
m2=m2+f(t);
end
s=(h/3)*(f(a)+f(b)+(4*m1)+(2*m2));
k=4*s;
fprintf('The required value of pi using Simpson 1/3 rule = %4.8f\n',k);
e=abs(pi-k);
disp('The error is =');
disp(e);

Output-

The number of intervals=6

The required value of pi using Simpson 1/3 rule = 3.14159178

The error is =

8.7265e-07

chick

The number of intervals=12

The required value of pi using Simpson 1/3 rule = 3.14159264

The error is =

1.3284e-08

The number of intervals=24

The required value of pi using Simpson 1/3 rule = 3.14159265

The error is =

2.0765e-10

(c) Simpson 3/8 Rule-

f=@(x)(1/(1+(x.^2)));
a=0;
b=1;
n=input('The number of intervals=');
h=(b-a)/n;
m1=0;
m2=0;
m3=0;
for i=1:3:n-1
t=a+(i*h);
m1=m1+f(t);
end
for i=2:3:n-2
t=a+(i*h);
m2=m2+f(t);
end
for i=3:3:n-3
t=a+(i*h);
m3=m3+f(t);
end
s=3*(h/8)*(f(a)+f(b)+(3*((m1)+(m2)))+(2*(m3)));
k=4*s;
fprintf('The required value of pi using Simpson 3/8 rule = %4.8f\n',k);
e=abs(pi-k);
disp('The error is =');
disp(e);

Output-

The number of intervals=6

The required value of pi using Simpson 3/8 rule = 2.69896050

The error is =

0.4426
The number of intervals=12

The required value of pi using Simpson 3/8 rule = 2.93781901

The error is =

0.2038

The number of intervals=24

The required value of pi using Simpson 3/8 rule = 3.04385510

The error is =

0.0977

Observation-

In the above three rules we see that error is inversely proportional to the no. of intervals.

Thus we can get the value of pi more accurate , if the no. of intervals become more larger.

Q(4).

Simpson 1/3 Rule-

f=@(x)(1/(1+(x.^2)));
a=0;
b=1;
n=2;
k=0;
while (abs(pi-k)>=5*(10.^-6))
h=(b-a)/n;
m1=0;
m2=0;
for i=1:2:n-1
t=a+(i*h);
m1=m1+f(t);
end
for i=2:2:n-2
t=a+(i*h);
m2=m2+f(t);
end
s=(h/3)*(f(a)+f(b)+(4*m1)+(2*m2));
k=4*s;
n=n+2;
end
fprintf('The required value of the integral by Simpson 1/3 rule =
%4.8f\n',s);

Output-
The required value of the integral by Simpson 1/3 rule = 0.78539795

Q(5).

Simpsons 1/3 rule-

f=@(x)((sqrt(1-(x.^2)))-x);
a=0;
b=1/sqrt(2);
n=2;
k=0;
while (abs(pi-k)>=5*(10.^-6))
h=(b-a)/n;
m1=0;
m2=0;
for i=1:2:n-1
t=a+(i*h);
m1=m1+f(t);
end
for i=2:2:n-2
t=a+(i*h);
m2=m2+f(t);
end
s=(h/3)*(f(a)+f(b)+(4*m1)+(2*m2));
k=8*s;
n=n+2;
end
fprintf('The required value of the integral by Simpson 1/3 rule =
%4.8f\n',s);

Output-

The required value of the integral by Simpson 1/3 rule = 0.39269866

Q(6).

(a) For qs no. 3-

% Gauss Legendre formula for neumerical integration for 2 points & 3 points %
f=@(x)(1/(1+(x.^2)));
a=0;
b=1;
F=@(t)f(((b-a)*t+(b+a))/2);
n=input('Enter the number of nodes:');
if n==2
w1=1;w2=1;
x1=1/sqrt(3);x2=-1/sqrt(3);
G2=((b-a)/2)*(w1*F(x1)+w2*F(x2));
k1=abs((pi/4)-G2);
fprintf('The value of the integration using gauss 2 point formula is =
%4.9f\n',G2);
fprintf('The error is = %4.8f',k1);
elseif n==3
w1=5/9;w2=8/9;w3=5/9;
x1=sqrt(3/5);x2=0;x3=-sqrt(3/5);
G3=((b-a)/2)*(w1*F(x1)+w2*F(x2)+w3*F(x3));
k2=abs((pi/4)-G3);
fprintf('The value of the integration using gauss 3 point formula is
:%4.9f\n',G3);
fprintf('The error is = %4.8f',k2);
end

Output-

Enter the number of nodes:2

The value of the integration using gauss 2 point formula is = 0.786885246

The error is = 0.00148708

Enter the number of nodes:3

The value of the integration using gauss 3 point formula is :0.785267035

The error is = 0.00013113

(b) For qs no. 5-

% Gauss Legendre formula for neumerical integration for 2 points & 3 points %
f=@(x)((sqrt(1-(x.^2)))-x);
a=0;
b=1/sqrt(2);
F=@(t)f(((b-a)*t+(b+a))/2);
n=input('Enter the number of nodes:');
if n==2
w1=1;w2=1;
x1=1/sqrt(3);x2=-1/sqrt(3);
G2=((b-a)/2)*(w1*F(x1)+w2*F(x2));
k1=abs((pi/8)-G2);
fprintf('The value of the integration using gauss 2 point formula is
:%4.9f\n',G2);
fprintf('The error is = %4.8f',k1);
elseif n==3
w1=5/9;w2=8/9;w3=5/9;
x1=sqrt(3/5);x2=0;x3=-sqrt(3/5);
G3=((b-a)/2)*(w1*F(x1)+w2*F(x2)+w3*F(x3));
k2=abs((pi/8)-G3);
fprintf('The value of the integration using gauss 3 point formula is
:%4.9f\n',G3);
fprintf('The error is = %4.8f',k2);
end

Output-

Enter the number of nodes:2

The value of the integration using gauss 2 point formula

is :0.393053552The error is = 0.00035447

Enter the number of nodes:3

The value of the integration using gauss 3 point formula

is :0.392714456The error is = 0.00001537

Observation-

The error is less in 3 points Gauss Legendry formula as compared to 2 points

Gauss legendryformula.hence we get more accurate value of integration in 3 points
Gauss legendry formula.