suggestions for improving GSL

Brook Milligan
Thu Mar 30 09:44:00 GMT 2000

Fabrice Rossi,

   Brook Milligan wrote:
   > - add multidimensional minimization to the set of optimization
   >   functions.  A candidate that I have found useful and easy to
   >   implement is described in Press, William H. and Teukolsky, Saul
   >   A. and Vetterling, William T. and Flannery, Brian P.
   >   1992. Numerical Recipies in C. The art of scientific computing.
   >   Second Edition.  Cambridge University Press, New York, New York.
   >   Section 10.4. Downhill simplex method in multidimensions. Pages
   >   408-412.  That might be a good start, though Press et al. also
   >   provide other ideas that could be added as well.  Note that if this
   >   is done, some capability (e.g., another function in addition to the
   >   one to be minimized) is required to specify whether points are
   >   within the domain of the function to be minimized; the published
   >   algorithm simply assumes that the domain extends throughout the
   >   entire N-dimensional space.

   I'm working on multidimensional minimization, currently for differentiable
   functions. The code is in the CVS, but is preliminary and lacks documentation.
   My next goal is to add new test functions and to start documenting.

Great!  I mentioned the above, though, because it doesn't require
derivatives, is fairly effective, and is easy to code.  Other
derivativeless methods might be better in some cases.  I hope you will
move on to include at least some such methods. 

It's great to hear that you are working on some multivariate
minimization, though.  Keep up the good work.


More information about the Gsl-discuss mailing list