On Mon, Oct 17, 2011 at 05:21:18PM -0400, Steven G. Johnson wrote:
> 
> On Oct 17, 2011, at 9:00 AM, Edd Barrett wrote:
> >
> >What I find interesting here is that an equality, say 'x[0] = 0' is
> >still subject to convergence; that is, you might get something very
> >close to 0, but not exactly. Is this right, or would this
> >suggest something is wrong? I would have thought that there is
> >absolutely no "slack" in an equality, as otherwise the solution is not
> >feasible.
> 
> In general, you can only converge asymptotically to a solution that
> obeys the equality constraints (limited eventually by roundoff
> errors).  In any finite number of steps they are only approximately
> met.
> 
> This is intrinsic to the algorithms, since you are in general asking
> them to solve an arbitrary nonlinear root-finding problem to satisfy
> nonlinear equality constraints.

Thanks for your reply. I understand.

-- 

Edd Barrett
Programming Languages and Systems Research
University of Kent
http://www.cs.kent.ac.uk/people/rpg/eb771/

_______________________________________________
NLopt-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/nlopt-discuss

Reply via email to