On Dec 5, 2011, at 1:07 AM, Mike Sumszyk wrote:
I have developped an algorithm in Matlab that uses the function
lsqcurvefit. This function implements the trust-region-reflective
algorithm in order to solve nonlinear curve-fitting (data-fitting)
problems in least-squares sense. I use it because it supports upper
and lower bounds constraints. According to the Matlab documentation:
[...]
I would like to know if there is an implementation of this algorithm
in the NLopt library? If no, is there some algorithm that would
perform similarly to the one I use in Matlab
There are plenty of Newton-like methods in NLopt that support bound
constraints (e.g. LBFGS, BOBYQA, ...), although it doesn't have the
specific algorithm you mention.
And, in general, NLopt can certainly be used for nonlinear curve
fitting with bound constraints (or with nonlinear constraints for that
matter).
The main caveat is that NLopt minimizes arbitrary nonlinear functions,
and unlike the lsqcurvefit function in Matlab it does not take
advantage of the special structure of a least-squares problem (http://www.mathworks.com/help/toolbox/optim/ug/brnoybu.html
). This means that the convergence may be somewhat slower than
lsqcurvefit, although your mileage may vary.
_______________________________________________
NLopt-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/nlopt-discuss
