On Jul 13, 2010, at 11:48 AM, Tom Fiddaman wrote:
This seems to be exactly the issue - changing the dimensionality
introduced
new variables with very different scales from the subset. I rescaled
all my
dimensions to a 0-1 hypercube and the problem goes away.
NLopt 2.2 will automatically do this rescaling for you, based on its
default initial step sizes in each dimension, which in turn are
derived from the bound constraints etcetera in each dimension. To
avoid automatic rescaling in version 2.2, you can just set all
dimensions to have the same initial step sizes via
nlopt_set_initial_step1.
Steven
_______________________________________________
NLopt-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/nlopt-discuss
