[R] Convex optimization in R?

[roger koenker]

I would be very wary of such approaches;  my experience is that MM is  
inferior
to the early affine-scaling versions of interior point algorithms for  
linear programming
problems, and modern implementations like the Mehrotra version of the  
primal dual
algorithm are much, much quicker and more reliable.   More general  
convex programming
is more delicate, and it is unlikely that methods that aren't that  
successful with LPs
improve their performance in more complex settings.  Something in R  
based on CVX
or Saunder's PDCO, or similar would be very welcome.  Meanwhile, as  
I've said
earlier on R-help, it is fairly convenient to link these options to R  
via R.matlab.

url:    www.econ.uiuc.edu/~roger            Roger Koenker
email    rkoenker at uiuc.edu            Department of Economics
vox:     217-333-4558                University of Illinois
fax:       217-244-6678                Champaign, IL 61820



On Sep 11, 2008, at 9:10 AM, Ravi Varadhan wrote:

>
> Ken Lange's MM `algorithm' is a possibility for these non-smooth,,  
> convex
> problems. It has been implemented in `constrOptim' for handling linear
> inequality constraints in the minimization of smooth objective  
> functions.  I
> have extended this to nonlinear inequalities.  It can be further  
> extended
> for convex functions, if one can come up with a smooth function that
> majorizes the convex objective function.  This can be easily done  
> for the
> absolute value function.
>
> Ravi.
>
>
> ----------------------------------------------------------------------------
> -------
>
> Ravi Varadhan, Ph.D.
>
> Assistant Professor, The Center on Aging and Health
>
> Division of Geriatric Medicine and Gerontology
>
> Johns Hopkins University
>
> Ph: (410) 502-2619
>
> Fax: (410) 614-9625
>
> Email: rvaradhan at jhmi.edu
>
> Webpage:  http://www.jhsph.edu/agingandhealth/People/Faculty/Varadhan.html
>
>
>
> ----------------------------------------------------------------------------
> --------
>
>
> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org 
> ] On
> Behalf Of Hans W. Borchers
> Sent: Thursday, September 11, 2008 7:19 AM
> To: r-help at stat.math.ethz.ch
> Subject: Re: [R] Convex optimization in R?
>
> Hesen Peng <hesen.peng <at> gmail.com> writes:
>
>>
>> Hi my R buddies,
>>
>> I'm trying to solve a specific group of convex optimization in R. The
>> admissible region is the inside and surface of a multi-dimensional
>> eclipse area and the goal function is the sum of absolution values of
>> the variables. Could any one please tell me whether there's a package
>> in R to do this? Thank you very much,
>
>
> To my knowledge there does not exist a designated R package for convex
> optimization. Also, in the Optimization task view the AMS nomenclature
> 90C25 for "Convex programming" (CP) is not mentioned.
>
> On the other hand, this task view may give you some ideas on how to  
> solve
> your problem with one of the available optimization packages.
> For instance, a problem including sums of absolute values can be  
> modeled as
> a linear program with mixed integer variables (MILP).
>
> There is a free module for 'disciplined' convex optimization, CVX,  
> that can
> be integrated with Matlab or Python. Hopefully, there will be a CVX R
> package in the future (as has been announced/promised).
>
> Hans Werner Borchers
> ABB Corporate Research
>
>
>> Best wishes,
>>
>> --
>> Hesen Peng
>> http://hesen.peng.googlepages.com/
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.