[SciPy-user] Questions about scipy.optimize.fmin_cobyla

dmitrey openopt@ukr....
Mon Jul 16 09:09:49 CDT 2007


fdu.xiaojf@gmail.com wrote:
> Can linear avoid the "math domain error" prolbem in cobyla ? In my
> function, there are items like log(xi*R*T/V) (R,T,V are positive
> variables too).
>   
do you mean "lincher" instead of "linear"?
no, at least for now (but some solvers, usually commercial, can handle 
NaN in objfun).

if R, T, V are constants I would replace

zi = log(xi*R*T/V) hence
xi = exp(xi)*V/(R*T)

hth, D

> Regards
>
>
>   



More information about the SciPy-user mailing list