Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.
|
|
|
|
Re: fmincon tolfun tolx and tolcon question/clarification
Posted:
Jan 28, 2013 10:48 AM
|
|
"Steven_Lord" <slord@mathworks.com> wrote in message <kdrhsr$qga$1@newscl01ah.mathworks.com>...
> "Jesse Sipple" <jesse.sipple@tufts.edu> wrote in message
> news:kdrfjg$h5j$1@newscl01ah.mathworks.com...
> > Alan_Weiss <aweiss@mathworks.com> wrote in message
> > <kdrca6$33a$1@newscl01ah.mathworks.com>...
> >> On 1/23/2013 9:08 PM, Jesse Sipple wrote:
>
> *snip*
>
> >> Take a look at the documentation of tolerances:
> >> http://www.mathworks.com/help/optim/ug/setting-options.html#brhv4_o-1
> >>
> >> Alan Weiss
> >> MATLAB mathematical toolbox documentation
> >
> > Thanks for responding so quickly!
> >
> > I've looked at that documentation before but I looked at it again and it
> > is still unclear.
> >
> > From studying that and looking at some more descriptions in optimtool, it
> > seems that TolCon is the constraint for the nonlinear constraints (c(x) >
> > TolCon or |ceq(x)| > TolCon).
>
> If I recall correctly (and I'm sure Alan will correct me if I don't) TolCon
> applies not just to the nonlinear constraints but also the linear
> constraints (A and b, Aeq and beq.) It may even apply to violations of the
> upper and lower bounds; I don't remember if those "count" as constraints.
>
*snip*
So I ran a test to see if TolCon had an effect on the optimization, the first round I left it as the default value of TolCon=1e-6 and recorded how long the optimization took and then changed only TolCon to 1e-16 and ran the same problem and the results took 7.5 times longer with slightly different values found.
But again, I have set no constraints (A, b, Aeq, Beq, nonlcon = []) other than lower and upper bounds (lb,ub) which my values are nowhere near so I'm unsure why changing TolCon is causing such a difference in the optimization.
|
|
|
|
