🚀 go-pugleaf

In article <haney.702746789@moonshine> haney@moonshine.llnl.gov (Scott W. Haney) writes:
>I am working on a Tokamak systems code that requires the use of a
>constrained minimizer in order to optimize the design.  Specifically,
>I need to minimize a function subject to a series of potentially non-linear
>equality and inequality constraints.  It would also be nice to be able
>to place direct bounds on variables.  I am currently using a routine
>named 'vmcon' but I'm not particularly happy with its performance. Does
>anyone know where I could get my hands on another package?  Thanks.
>
>   Scott
>
>FYI, the problem consists of roughly 20 equality constraints, 15 inequality
>constraints, and 40 variables.

  NPSOL is very good.  It is the Gill, Murray, Saunders and Wright code.

  Contact the Office of Technology Licensing at Stanford University at
  (415) 723-0651 for more information.  They will send manuals and
  license agreements on request.  Academic licenses are cheap (a $300
  one-time fee last time I got one).  I guess that includes goverment
  labs, but I don't know.  This is also routine E04UCF in the NAG Fortran
  library if you've got that.

  FSQP is some software that I got but haven't had a chance to try.  Its
  main difference from NPSOL is that it only evaluates the objective
  function and constraints at feasible points (with NPSOL the nonlinear
  constraints are only satisfied at the solution).

  It is available from the authors without charge to academic users but
  not freely redistributable.  E-mail andre@src.umd.edu for info.

Hope this helps.  I've used NPSOL a lot and it works great.  Just out of
curiosity, what is "vmcon" and where does it come from?

-- 
Charles Geyer
School of Statistics
University of Minnesota
charlie@umnstat.stat.umn.edu