22.01.2014 Views

download searchable PDF of Circuit Design book - IEEE Global ...

download searchable PDF of Circuit Design book - IEEE Global ...

download searchable PDF of Circuit Design book - IEEE Global ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

148 Gradient Optimization<br />

The output for the Rosenbrock problem, starting at Xl = I and x,= -I, is<br />

shown in Table 5.4 (see Figure 5.24). The data show that ITN = 21 iterations<br />

(linear search directions) and IFN =61 function and gradient evaluations are<br />

required to locate the global minimum at x=(I,I)T to at least 0.00001<br />

accuracy in each variable.<br />

The reader should run this example to observe the effects <strong>of</strong> several<br />

changes. Several new starting points should be tried. The accuracy set in line<br />

160 can be reduced. The number <strong>of</strong> variables can be set to 20 instead <strong>of</strong> 2 by<br />

inputting the latter number when asked and setting all but the first two<br />

variables to an arbitrary number, e.g., O. This will illustrate how much <strong>of</strong> the<br />

computing time is in search overhead, because the full 20 variables will be<br />

treated by the Fletcher-Reeves algorithm even though only the first two<br />

determine the problem defined in subroutine 1000. It is informative to add the<br />

statement 392 Z = O. This causes the search to be <strong>of</strong> steepest descent at all<br />

1.0<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

a.'<br />

0.3<br />

0.2<br />

Fletcher-Reeves<br />

Optimum value <strong>of</strong> x~'<br />

, ,,,,,, /<br />

, , /<br />

/<br />

/.~.<br />

, j<br />

-0.1 \ 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Xl<br />

-0.1\<br />

-0.2<br />

-0.5<br />

-0.6<br />

Steepest descent<br />

-0.7<br />

-0.8<br />

Figure 5.24.<br />

-0.9<br />

-1.0<br />

Starting value <strong>of</strong> x --:lIo.<br />

Trajectories in the x plane for the Fletcher-Reeves and steepest-descent algorithms.

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

Saved successfully!

Ooh no, something went wrong!