Hi Christian,
The best way to handle master-only columns is explicitly. Unfortunately, I
never got around to finish coding that up in the framework. It is on my
longer-term ToDo List.
Currently, the way it works in DIP is by using dummy blocks for each
master-only column. These are constructed relatively efficiently, so it is
not as bad as it sounds. Although it does add a convexity constraint for
each block - which can bog down the master problem if there are many
master-only columns.
The MILPBlock application will create, identify and manage these for you. If
you are building your own application, you have to let the framework know
which columns are master-only by creating blocks (just like constraint
blocks) and then flagging them as master only. I don't think I have a simple
example of this written up. But, you can see how to do it in the MILPBlock
application. See MILPBlock_DecompApp.cpp createModelMasterOnlys( ).
for(vit = masterOnlyCols.begin(); vit != masterOnlyCols.end(); vit++){
i = *vit;
DecompConstraintSet * model = new DecompConstraintSet();
model->m_masterOnly = true;
model->m_masterOnlyIndex = i;
model->m_masterOnlyLB = colLB[i];
model->m_masterOnlyUB = colUB[i];
//0=cont, 1=integer
model->m_masterOnlyIsInt =
(integerVars && integerVars[i]) ? true : false;
//...
setModelRelax(model,
"master_only" + UtilIntToStr(i), nBlocks);
nBlocks++;
//...
Let me know if you have any questions.
Thanks,
Matt
On Fri, Oct 7, 2011 at 5:58 AM, Christian Plum <chrplum at gmail.com> wrote:
> Hi,
>
> Im working on price and cut model containing Master-only Columns, which are
> not appearing in any block. This is briefly mentioned in DecompAlgo.cpp Ln
> 134:
>
> //---
> //--- sanity check that the union of active columns in blocks
> //--- should cover all columns in core - if not, these are
> 'master-only'
> //--- columns which can be dealt with using either LD or the using
> the
> //--- ideas of Rob Pratt discussion (9/27/09), or defined explicitly
> //--- by user
> //---
>
> Could this refer to creating a dummy block for these columns ? Or how is
> this best handled ?
>
> Brgds, Christian
>
>
>
> _______________________________________________
> Dip mailing list
> Dip at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/dip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/dip/attachments/20111007/c95fdf6f/attachment.html>