15.05.2015 Views

Introduction to Unconstrained Optimization - Scilab

Introduction to Unconstrained Optimization - Scilab

Introduction to Unconstrained Optimization - Scilab

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

3000<br />

2500<br />

2000<br />

Z<br />

1500<br />

1000<br />

500<br />

0<br />

2.0<br />

1.5<br />

1.0<br />

0.5<br />

Y 0.5<br />

0.0<br />

0.0<br />

-0.5 X<br />

-0.5 -1.0<br />

-1.5<br />

-1.0 -2.0<br />

1.0<br />

1.5<br />

2.0<br />

Figure 4: Three dimensionnal plot of Rosenbrock’s function.<br />

data as input arguments.<br />

function f = rosenbrockS ( x1 , x2 )<br />

f = rosenbrock ( [ x1 x2 ] )<br />

endfunction<br />

The following statements allows <strong>to</strong> produce the plot presented in figure 4. Notice<br />

that we must transpose the output of feval in order <strong>to</strong> feed surf.<br />

x = linspace ( -2 , 2 , 20 );<br />

y = linspace ( -1 , 2 , 20 );<br />

Z = feval ( x , y , rosenbrockS );<br />

surf (x,y,Z ’)<br />

h = gcf ();<br />

cmap = graycolormap (10);<br />

h. color_map = cmap ;<br />

On one hand, the 3D plot seems <strong>to</strong> be more informative than the con<strong>to</strong>ur plot.<br />

On the other hand, we see that the level sets of the con<strong>to</strong>ur plot follow the curve<br />

quadratic curve x 2 = x 2 1, as expected from the function definition. The surface has<br />

the shape of a valley, which minimum is at x ⋆ = (1, 1) T . The whole picture has the<br />

form of a banana, which explains why some demonstrations of this test case present<br />

it as the banana function. As a matter of fact, the con<strong>to</strong>ur plot is often much more<br />

simple <strong>to</strong> analyze than a 3d surface plot. This is the reason why it is used more<br />

11

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!