[SciPy-user] optimization with bounded parameter values

[John Hunter]

>>>>> "Michael" == Michael Sorich <mike_sorich at hotmail.com> writes:

    Michael> Hi All, I am trying to solve a nonlinear regression
    Michael> problem with 4 inputs and 8 parameters. I have used the
    Michael> leastsq function, however, the parameter values returned
    Michael> are not realistic and thus I would like to constrain the
    Michael> parameters to be within certain bounds. Is it possible to
    Michael> do this in scipy? If not, I would welcome suggestions of
    Michael> other free code (preferably in python) or software that
    Michael> can.

qoca looks nice for a bounded optimizer; you can do linear arithmetic
constraints, and there is a python interface

  http://www.csse.monash.edu.au/projects/qoca/

Last time I tried to build it, several months ago, I ran into some
problems with an out-of-date prerequisite ch2xx and got side tracked
and so haven't succeeded in building and testing it.  If you use
debian, however, there is a package.

Let me know if you succeed.

John Hunter