[Scilab-users] 'fsolver'

[fujimoto2005]

Hi, motterlet and Steer

Thanks a lot your helps.

Unfortunately 'lsqrsolve' did't give the smallest initial point for my
function.
The local minimum of f^2+a*x^2 is attained at x s.t. f(x)=-a*x.
So if a is small such x is a neighborhood of x s.t. f(x)=0.
But it is not necessarily of x which is the smallest one.
Probably my function is not well-behaved as like cos(x) so that it fail.

Now I get an awkward method.
I find the first x(i) s.t. f(x(i))>0 and f(x(i-1))<0 where x(i)=x0+i*dx and
x0 is some constant which is smaller than smallest solution of f(x)=0.
Then I modify f(x) to new function fnew(x) as
f(x)=f(x(i-1))+[f(x(i))-f(x(i-1))]/[x(i)-x(i-1)]*[x-x(i)] for x>=x(i) and
fnew(x)=f(x) for x<=x(i).
Using fsolv(x(i-1),fnew) gives the smallest solution of f(x)=0.
With your helps I could get a practical solution.
Thanks again.

Best regards.




--
View this message in context: http://mailinglists.scilab.org/fsolver-tp4033340p4033350.html
Sent from the Scilab users - Mailing Lists Archives mailing list archive at Nabble.com.