[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: add optimization function -> fmincon
From: |
Juan Pablo Carbajal |
Subject: |
Re: add optimization function -> fmincon |
Date: |
Fri, 20 Jan 2012 14:46:50 +0100 |
On Fri, Jan 20, 2012 at 2:11 PM, Ben Abbott <address@hidden> wrote:
> On the IRC yesterday, it was suggested that ML's fmincon was sqp in disguise.
>
> Are there any optimizations experts on the list that have an opinion on the
> idea of wrapping an m-file around sqp to provide an fmincon.m that is
> compatible with Matlab's?
>
> I've opened a feature request on the tracker.
>
> https://savannah.gnu.org/bugs/?35333
>
> Ben
>
Hi Ben,
By answering this question I do not want to give the impression I am
an expert, however I collaborate with some people who uses fmincon in
our joint code. As far as I can tell (based on the way they emulate my
call to Octave's sqp with fmincon), fmincon has some other algorithms
in addition to sqp. The call usually looks like this (in my colleagues
code)
nonLinConst = @(p_) createNonLinConst(p_, ineqConst, eqConst);
optFmincon = optimset('Algorithm','sqp','Display','iter');
[p,fval,exitflag,output] = fmincon(objFun,p0,[],[],[],...
[],[],[], nonLinConst, optFmincon);
emulating my Octave code
[param,fval,exitflag,iter,nf,lambda] = ...
sqp (p0,objFun,eqConst,ineqConst,...
[],[],[],[]);
Where ineqConst and eqConst stands for the inequalities and equalities
constraint handles, respectively.
Therefore, I see no reason not to provide fmincon as a different
interface for sqp. Although I do not see more fruits than
compatibility.
Also, it sounds like a boring job since the interface is quite different.
Let me know if I can help you somehow.
--
M. Sc. Juan Pablo Carbajal
-----
PhD Student
University of Zürich
http://ailab.ifi.uzh.ch/carbajal/