Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
46 views

Classnotes - Coursework of Optimization

The document discusses solving optimization problems using various numerical methods including univariate, Powell's method, and Cauchy's method. For problem A3, the first equation is solved in 15 iterations using univariate method, resulting in a final solution of [1.0078; 0.9922]. The second equation is solved in 11 iterations, giving a solution of [1.9959; 0.9959]. Powell's method solves the first equation in 9 iterations with a solution of [0.7097; 1.4354] and the second equation in 8 iterations with a solution of [1.9848; 1.0042]. Cauchy's method is applied to an example function, iterating up to 30 times to minimize

Uploaded by

Hudzaifa Ahnaf
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
46 views

Classnotes - Coursework of Optimization

The document discusses solving optimization problems using various numerical methods including univariate, Powell's method, and Cauchy's method. For problem A3, the first equation is solved in 15 iterations using univariate method, resulting in a final solution of [1.0078; 0.9922]. The second equation is solved in 11 iterations, giving a solution of [1.9959; 0.9959]. Powell's method solves the first equation in 9 iterations with a solution of [0.7097; 1.4354] and the second equation in 8 iterations with a solution of [1.9848; 1.0042]. Cauchy's method is applied to an example function, iterating up to 30 times to minimize

Uploaded by

Hudzaifa Ahnaf
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 14

Hudzaifa Dhiaul Ahnaf

02311940000102
Optimization C
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C

Problem A3

Univariate

Figure 1, First Equation

Figure 2, Second Equation


Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C

Powells

Figure 3, First Equation

Figure 4, Second Equation


Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C

Problem A4

Univariate

Second Function
%%
clear all;
clc ;
%% Level Sets

[X,Y]=meshgrid(-10:5);
z=f(X,Y);
contour(X,Y,z,[1:6])
%% First State
syms x1 x2 xl
eps = 0.01;
i = 1;
n = 2;
S = eye(2);
err = 1 ;

%%
x = [0;0];
% f = symfun(2*x1^2 + x2^2 + 2*x1*x2 - 6*x1 - 4*x2 , [x1 , x2])
f = symfun(3*x1^2 + x2^2 - 2*x1*x2 - 10*x1 + 2*x2 , [x1 , x2])

%% Iteration
while err > 0
x_temp = x(:,i);
if mod(i,2) == 1 % Odd
s = S(:,1)
else
s = S(:,2) % even
end

fval = f(x_temp(1),x_temp(2))
xp_temp = x_temp + eps*s
xn_temp = x_temp - eps*s
fval_p = f(xp_temp(1),xp_temp(2))
fval_n = double(f(xn_temp(1),xn_temp(2)))

if fval_p < fval


s = s;
elseif fval_n < fval
s = -s;
else
break;
end

xopt_temp = x_temp + xl*s


fval_opt = f(xopt_temp(1),xopt_temp(2))
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C

grad_fopt = gradient(fval_opt)
lambda_opt = solve(grad_fopt)

x(:,i+1) = x_temp + lambda_opt*s

i = i+1

err = abs(((f(x(1,i),x(2,i)))-(f(x(1,i-1),x(2,i-1))))/...
(f(x(1,i-1),x(2,i-1)))) %determine the convergence criterion

end

axis square
x1 = linspace(-2.5,2.5);
x2 = linspace(-2.5,2.5);
[A,B] = meshgrid(x1,x2);
f_fig = f(A,B);
levels = 10:10:350;
figure(1), contour(x1,x2,f_fig,levels,...
'LineWidth',1.2), colorbar
hold on;
plot(x(1,:),x(2,:))

With unicariate, the first equation will be solved within the 15th iteration, final X is

[1.0078 ; 0.9922]

With univariate, the second equation will be solved within 11st iteration, will get final solution of X :
[1.9959; 0.9959]

Powell’s
%%
clear all;
clc ;
%% First State
syms x1 x2 xl
eps = 0.1;
i = 1;
n = 2;
S = [1 0 0 0 0 0; 0 1 0 0 0 0]
err = 1 ;
m = 1

%% Obj fun
x = [0;0];
f = symfun(2*x1^2 + x2^2 + 2*x1*x2 - 6*x1 - 4*x2 , [x1 , x2])
% f = symfun(3*x1^2 + x2^2 - 2*x1*x2 - 10*x1 + 2*x2 , [x1 , x2])
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C

%% Iteration
while err > 0.001
x_temp = x(:,i);

if i == 1
s = S(:,2)

elseif mod(i,3) == 2
s = S(:,(i-(2*m-1)))

elseif mod(i,3) == 0
s = S(:,i-(2*m-1))

elseif mod(i,3) == 1
s = x(:,i) - S(:,(i-(2*m)));
S(:,i-1) = s
m = m + 1;

end

fval = double(f(x_temp(1),x_temp(2)))
xp_temp = x_temp + eps*s
xn_temp = x_temp - eps*s
fval_p = double(f(xp_temp(1),xp_temp(2)))
fval_n = double(f(xn_temp(1),xn_temp(2)))

if fval_p < fval


s = s;
elseif fval_n < fval
s = -s;
else
break;
end

xopt_temp = x_temp + xl*s


fval_opt = f(xopt_temp(1),xopt_temp(2))
grad_fopt = gradient(fval_opt)
lambda_opt = solve(grad_fopt)

x(:,i+1) = x_temp + lambda_opt*s


i = i+1
if i > 2
err = abs(((f(x(1,i),x(2,i)))-(f(x(1,i-1),x(2,i-1))))/...
(f(x(1,i-1),x(2,i-1)))) %determine the convergence criterion
end
end
%% Plot
% for i = 1:20
% % while fval = 0/error > threshold
%
% % S = ? POWELL, UPDATE S
% % error =
% % menyipman x, x diplot menjadi zig zag
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C

% % xk yk ini disimpan, diplot 100 nilai Jk kemudian diplot, k 1 sampai


% % 100
% end

axis square
x1 = linspace(-50,50);
x2 = linspace(-50,50);
[A,B] = meshgrid(x1,x2);
f_fig = f(A,B);
levels = 10:10:350;
asd = figure(1), contour(x1,x2,f_fig,levels,...
'LineWidth',1.2), colorbar
hold on;
plot(x(1,:),x(2,:))

With powell’s the first equation solved within 9th iteration, and resulting in x = [0. 7097 ; 1. 4354]

And second equation solved within 8th iteration, resulting in x = [1. 9848 ; 1.0042]

Problem A5

When we use different starting point, the start line will change differently depending on the how far the
optimized point is, when we change the epsilon or the step, increasing to 0.000 … 01 wil get the
optimized value more time because the iteration will increasing or decreasing slowly. Or reducing the
epsilon to 1 will get the result more fast.

Problem B1
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C

Problem B2

Cauchy

Problem B3
%%
clear all;
clc ;

%% First State
syms x y lambda
eps = 0.01;
n = 2;
err = 1 ;
G = [0 ; 0];
i = 1

%%
f = symfun( 100*y^2 + 100*x^4 - 200*y*x^2 + x^2 - 2*x + 1 , [x y]);
X0 = [-0.75;1];

%% b) Direction S
while err > 0.01
gradxx = gradient(f,x) %diff the f by x
gradyy = gradient(f,y)

G(1,1) = gradxx(X0(1,i),X0(2,i)) %assign gradient val to G


G(2,1) = gradyy(X0(1,i),X0(2,i))

S = -G %assign G to S by multiply by -
Hudzaifa Dhiaul Ahnaf
02311940000102
Optimization C

xopt = X0(:,i) + lambda*S %The X to be assigned to f


fopt = f(xopt(1),xopt(2)) %The f that assigned from x opt
gradopt = gradient(fopt) %The value of df/dlambda
lambdaopt = double(real(solve(gradopt, lambda))) %Solving gradient equal to 0

X0(:,i+1) = X0(:,i) + lambdaopt(1)*S

if i > 2
err = double(abs((f(X0(1,i),X0(2,i))-f(X0(1,i-1),X0(2,i-1)))/(f(X0(1,i-
1),X0(2,i-1)))))
%determine the convergence criterion
end

i = i+1

end

%%
axis square
x = linspace(-2.5,2.5);
y = linspace(-2.5,5);
[A,B] = meshgrid(x,y);
f_fig = f(A,B);
levels = 10:10:350;
figure(1), contour(x,y,f_fig,levels,...
'LineWidth',1.2), colorbar
hold on;
plot(X0(1,:),X0(2,:))

will get the iteration up to 30 iteration.

You might also like