Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
about exitflag value of fmincon
932 views
Skip to first unread message
Ashley
Jul 13, 2012, 10:43:32 PM7/13/12
to
Hi?everyone
I‘m thinking about this question for quite a long time
we all know that we got good and reliable results when run some optimization functions such as fmincon ,we got 1 or at least a positive number of exitflag
but what if we got a negative number?Does it mean that these results are not reliable?But the fact is that sometimes these results seem to be much better because they are more approximate to the experiment results.
I don't know if we should lower down the tolerance of some constraints or just ignore the negative exitflag.
hope for help?thank you
I‘m thinking about this question for quite a long time
we all know that we got good and reliable results when run some optimization functions such as fmincon ,we got 1 or at least a positive number of exitflag
but what if we got a negative number?Does it mean that these results are not reliable?But the fact is that sometimes these results seem to be much better because they are more approximate to the experiment results.
I don't know if we should lower down the tolerance of some constraints or just ignore the negative exitflag.
hope for help?thank you
Sargondjani
Jul 14, 2012, 5:13:07 AM7/14/12
to
would always want a positive exit flag... i think if you have a negative exit flag there is a problem that you should solve. especially if those results are better than with the positive exit flag (this hints that some is wrong indeed)
but i think ppl will need more details to give proper advice what to do... what specific exit flag do you get? (and why?) And maybe describe your problem (or some code of the optimization)
but i think ppl will need more details to give proper advice what to do... what specific exit flag do you get? (and why?) And maybe describe your problem (or some code of the optimization)
Alan_Weiss
Jul 16, 2012, 8:12:44 AM7/16/12
to
http://www.mathworks.com/help/toolbox/optim/ug/fmincon.html#f469794
In particular, you see that when the exit flag is negative, usually the
"solution" is infeasible, or fmincon ran out of iterations or function
evaluations, so the "solution" cannot be considered reliable.
If you have a relatively recent version of Optimization Toolbox, the
exit message contains a link describing fmincon's estimate of the
quality of the result.
Alan Weiss
MATLAB mathematical toolbox documentation
0 new messages