On Jul 23, 2012, at 3:00 AM, ben Thelen wrote:
>
> I'd observed that upon failure (error raised) in the objective function the
> optimizer would somehow register this (presumably the same as returning -inf
> to a max. objective) and continue to optimize.
I can't replicate this behavior, nor is it what is supposed to be happening.
As documented in the NLopt manual, raising an exception in your objective (or
constraint) function is equivalent to a forced stop: it halts the optimization
and raises the exception to the code that called NLopt:
http://jdj.mit.edu/wiki/index.php/NLopt_Python_Reference#Forced_termination
For example, if I modify the Python example in the NLopt tutorial
(http://jdj.mit.edu/wiki/index.php/NLopt_Tutorial#Example_in_Python) to raise
an exception in the objective function:
def myfunc(x, grad):
raise Exception('spam','eggs')
if grad.size > 0:
grad[0] = 0.0
grad[1] = 0.5 / sqrt(x[1])
return sqrt(x[1])
then it halts the optimization and raises this exception to the caller, as can
be verified by wrapping the opt.optimize call in try: ... except.
Can you modify the tutorial code in a way that reproduces your problem?
Steven_______________________________________________
NLopt-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/nlopt-discuss
