[GSoC-PortA] ROI status

Stefan Theussl stefan.theussl at gmail.com
Thu Nov 28 15:07:28 CET 2013 

Hi,

first of all let me wish you a happy thanksgiving!

Back to the topic:

I just submitted ROI and plugins to CRAN. They will soon distribute through
the network. I managed to fix the variable bounds issue with the quadprog
plugin. I simply had to translate Inf to .Machine$double.xmax in one of the
plugin routines. that was all.

I hope everything checks fine in your test battery now. At least the one
example you provided and a couple of examples on my side do. Otherwise,
please just drop me an email. I am happy to improve the package and provide
releases in a higher frequency from now on. I and colleagues here at the
bank started to use ROI in backtests as well, so it should become well
tested soon. However, I advice you to cross check results. The ROI is still
experimental.

User visible changes:
o "bounds" are defined using a constructor called V_bound(). Corresponding
class has a print method now.
o in many cases the return value has an additional slot in attribute meta
(see attr( result, "meta" )$solver_return_object in the example you
provided) which includes the solution in the original solver format. More
information about the solution is provided.
o improved print methods
o a lot of changes behind the scenes

o (currently not exported) compute the solvers capable to solve a given OP
named x: ROI:::get_solver_methods( ROI:::OP_signature(x) ).

Best,
st



On Thu, Nov 28, 2013 at 12:23 AM, Stefan Theussl
<stefan.theussl at gmail.com>wrote:

> Hi,
>
> thanks for testing. Indeed one needs the latest version of ROI, otherwise
> the plugins might not work as expected (I added corresponding version
> dependencies to the DESCRIPTION files of the plugins).
>
> Since R-Forge does not understand roxygen2, package documentation adn
> NAMESPACE are not built. This is why ROI must be built from SVN directly. I
> send you the package attached.
>
> Also, with the latest version variable bounds are constructed using a
> specific constructor function. see the modified version of your script. One
> issue with the quadprog plugin still remains, namely that the unbounded QP
> as defined in your script is not solvable. Looking through the code I can
> imagine where this might happen (ROI:::as.no_V_bound_OP()). I will check
> this in the morning.
>
> cheers,
> st
>
>
> On Wed, Nov 27, 2013 at 11:22 PM, Ross Bennett <rossbennett34 at gmail.com>wrote:
>
>> All,
>>
>> I installed ROI.plugin.quadprog from R-forge successfully. I tried
>> installing ROI from R-forge, but got the following error.
>> * installing to library
>> ‘/Library/Frameworks/R.framework/Versions/3.0/Resources/library’
>> * installing *source* package ‘ROI’ ...
>> ERROR: a 'NAMESPACE' file is required
>>
>> Also, I am getting an NA for the unconstrained global minimum variance
>> solution. This used to work the way I am doing it with -Inf and Inf for the
>> lower and upper bounds. See attached script for a self-contained example of
>> how I implement this using ROI.
>>
>> Let me know if this should be implemented differently.
>>
>> Thanks,
>> Ross
>>
>>
>>
>> On Wed, Nov 27, 2013 at 1:05 PM, Ross Bennett <rossbennett34 at gmail.com>wrote:
>>
>>> This is great and I will go ahead and test. I was using solve.QP and
>>> Rglpk directly for a few of the objectives as a workaround so I will switch
>>> these back over to ROI.
>>>
>>> I'll report back if I run into any errors.
>>>
>>> Ross
>>>
>>>
>>> On Wed, Nov 27, 2013 at 12:25 PM, Brian G. Peterson <brian at braverock.com
>>> > wrote:
>>>
>>>> Stefan,
>>>>
>>>> Thank you.  We'll certainly get the latest from R-Forge and test.  We
>>>> have a reasonably complete set of demo and test scripts at this point,
>>>> which hopefully would uncover any problems.
>>>>
>>>> Regards,
>>>>
>>>> Brian
>>>>
>>>> n 11/26/2013 07:29 AM, Stefan Theussl wrote:
>>>>
>>>>> Hi Brian,
>>>>>
>>>>> I recently improved ROI and their plugins. I will submit them to CRAN
>>>>> today or tomorrow. It would be great if you checked the current version
>>>>> of ROI and plugins (glpk, symphony, quadprog) with your framework. It
>>>>> would help me a lot in testing it.
>>>>>
>>>>> Anyway, I am about to implement further tests right now. But as always:
>>>>> the more the better ...
>>>>>
>>>>> Best,
>>>>> st
>>>>>
>>>>>
>>>>> On Tue, Oct 29, 2013 at 8:36 PM, Stefan Theussl
>>>>> <stefan.theussl at gmail.com <mailto:stefan.theussl at gmail.com>> wrote:
>>>>>
>>>>>     Hello Brian,
>>>>>
>>>>>     thanks, all good here. How are you?
>>>>>
>>>>>     ROI is definitely not to be abandoned. Nevertheless, I have to
>>>>> admit
>>>>>     that I have not worked on it for a longer time due to time
>>>>>     constraints. Only an update of Rglpk made it to CRAN some time ago.
>>>>>
>>>>>     Anyway, I just put the ROI plugin to quadprog on top of my TODO
>>>>>     list. I am looking forward to using PortfolioAnalytics myself and I
>>>>>     appreciate the plan, that ROI might be part of it.
>>>>>
>>>>>     Best,
>>>>>     Stefan
>>>>>
>>>>>
>>>>>     On Sun, Oct 27, 2013 at 8:11 PM, Brian G. Peterson
>>>>>
>>>>>     <brian at braverock.com <mailto:brian at braverock.com>> wrote:
>>>>>
>>>>>         Stefan,
>>>>>
>>>>>         I hope things are well with you.
>>>>>
>>>>>         I'm writing to ask what your plans are longer term for ROI.
>>>>>
>>>>>         Kurt indicated that you are still working on it, but there have
>>>>>         been almost no changes in the last year.
>>>>>
>>>>>         Specifically, the change that we worked on with you a year ago
>>>>>         for the quadprog plugin has not made it to CRAN.
>>>>>
>>>>>         More generally, if ROI is to be abandoned, I need to start
>>>>>         planning for significant changes to PortfolioAnalytics before I
>>>>>         send it to CRAN, as we now use ROI almost exclusively for
>>>>>         solvers that are supported by ROI.
>>>>>
>>>>>         I think the framework is valuable, and I hope that you can move
>>>>>         the quadprog fix to CRAN, and let me know how the package is to
>>>>>         be maintained long-term.
>>>>>
>>>>>         Best Regards,
>>>>>
>>>>>         Brian
>>>>>
>>>>>         --
>>>>>         Brian G. Peterson
>>>>>         http://braverock.com/brian/
>>>>>         Ph: 773-459-4973 <tel:773-459-4973>
>>>>>         IM: bgpbraverock
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>> --
>>>> Brian G. Peterson
>>>> http://braverock.com/brian/
>>>> Ph: 773-459-4973
>>>> IM: bgpbraverock
>>>> _______________________________________________
>>>> GSoC-PortA mailing list
>>>> GSoC-PortA at lists.r-forge.r-project.org
>>>> http://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/gsoc-porta
>>>>
>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/gsoc-porta/attachments/20131128/add92e92/attachment.html>

More information about the GSoC-PortA mailing list