[Dip] SmallIP Example and Implicity Polyhedra

Matthew Galati Matthew.Galati at sas.com
Sun Sep 5 10:18:05 EDT 2010


To help illustrate -- I will create several variants of the GAP example (sparse build, dense build, built-in MILP oracle, user-defined oracle, etc) to show the different use cases. I'll let you know when that is done.

Thanks,
Matt


From: kippmartin71 at yahoo.com [mailto:kippmartin71 at yahoo.com]
Sent: Saturday, September 04, 2010 2:21 AM
To: Matthew Galati
Cc: dip at list.coin-or.org
Subject: SmallIP Example and Implicity Polyhedra

Hi:

In the Dip/examples/SmallIP

the use of

   setModelRelax(NULL);

is illustrated, presumably for implicitly defined polyhedra. But in this example there is only one block. When using

   setModelRelax(NULL);

Hi:

In the Dip/examples/SmallIP

the use of

   setModelRelax(NULL);

is illustrated, presumably for implicitly defined polyhedra. But in this example there is only one block. When using

   setModelRelax(NULL);

how do you specify the block?

I need to solve a problem that is block angular (more than one block) and not give an explicit description of the block polyhedra. I am still really confused on how to do this. Is the following sufficient?

1) For each block create a DecompConstraintSet, i.e.

modelRelax = new DecompConstraintSet();


2) For modelRelax in each block specify the columns that define the block using

modelRelax->activeColumns.push_back( i)

3) For each block call the method

modelRelax->fixNonActiveColumns();

4)  For each block call


 setModelRelax(NULL);

But how can this work since the setModelRelax does not know the block name or number?  I have studied the examples, but it seems like in every example the constraint matrix is specified for the blocks and I don't see an example of setModelRelax() with more than one block.  I just cannot figure out what the sufficient steps are for this case.

Thanks

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/dip/attachments/20100905/b7dbb3de/attachment.html 


More information about the Dip mailing list