262 Nonlinear Programming I: One-Dimensional Minimization Methods
Example 5.6 Find the minimum off=x(x− 1 .5) in the interval (0.0, 1.0) to within
10% of the exact value.
SOLUTION If the middle point of the final interval of uncertainty is taken as the
optimum point, the specified accuracy can be achieved if
1
2
Ln≤
L 0
10
or
(
1
2
)(n − 1 )/ 2
L 0 ≤
L 0
5
(E 1 )
SinceL 0 = , Eq. 1 (E 1 ) ivesg
1
2 (n^ −^1 )/^2
≤
1
5
or 2(n^ −^1 )/^2 ≥ 5 (E 2 )
Sincenhas to be odd, inequality (E 2 ) gives the minimum permissible value ofnas 7.
With this value ofn=7, the search is conducted as follows. The first three experiments
are placed at one-fourth points of the intervalL 0 = [a= 0 ,b=1] as
x 1 = 0. 25 , f 1 = 0. 25 (− 1. 25 )= − 0. 3125
x 0 = 0. 50 , f 0 = 0. 50 (− 1. 00 )= − 0. 5000
x 2 = 0. 75 , f 2 = 0. 75 (− 0. 75 )= − 0. 5625
Sincef 1 >f 0 >f 2 , we delete the interval (a, x 0 ) =(0.0, 0.5), labelx 2 andx 0 as the
newx 0 andaso thata= 0 .5,x 0 = 0 .75,andb= 1 .0. By dividing the new interval of
uncertainty,L 3 = 0.5, 1.0) into four equal parts, we obtain(
x 1 = 0. 625 , f 1 = 0. 625 (− 0. 875 )= − 0. 546875
x 0 = 0. 750 , f 0 = 0. 750 (− 0. 750 )= − 0. 562500
x 2 = 0. 875 , f 2 = 0. 875 (− 0. 625 )= − 0. 546875
Sincef 1 >f 0 andf 2 >f 0 , we delete both the intervals (a, x 1 ) nd (a x 2 , b) ,and label
x 1 ,x 0 , andx 2 as the newa,x 0 , and b,respectively. Thus the new interval of uncer-
tainty will beL 5 = (0.625,0.875). Next, this interval is divided into four equal parts
to obtain
x 1 = 0. 6875 , f 1 = 0. 6875 (− 0. 8125 )= − 0. 558594
x 0 = 0. 75 , f 0 = 0. 75 (− 0. 75 )= − 0. 5625
x 2 = 0. 8125 , f 2 = 0. 8125 (− 0. 6875 )= − 0. 558594
Again we note thatf 1 >f 0 andf 2 >f 0 and hence we delete both the intervals (a, x 1 )
and (x 2 , b) to obtain the new interval of uncertainty asL 7 = 0.6875, 0.8125). By(
taking the middle point of this interval (L 7 ) s optimum, we obtaina
xopt≈ 0. 7 5 and fopt≈ − 0. 5625
(This solution happens to be the exact solution in this case.)