On 9/19/2014 11:32 AM, Prof J C Nash (U30A) wrote:
One choice is to add a penalty to the objective to enforce the constraint(s) along with bounds to keep the parameters from going wild. This generally works reasonably well. Sometimes it helps to run just a few iterations with a big penalty scale to force the parameters into a feasible region, though a lot depends on the particular problem in my experience, with some being straightforward and others needing a lot of fiddle. I suspect a math programming approach is overkill, though R does have some packages for that. Your mileage may vary. Note that L-BFGS-B used by R is a version for which Nocedal et al. reported a bug in 2011 and provided a new Fortran code. I've recently put up an implementation of the new code on r-forge under the optimizer project. Still testing, but I think it's OK. You could also use Rvmmin that has bounds, or nmkb from dfoptim (though you cannot start on bounds). Best, JN
Thanks very much. I've found that for most of my current problems, optimx is working well, but there are trade-offs. Good to have multiple options in the toolkit.
Cheers... ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
