[Osi] Fwd: crossover step

[Matthew Galati]

Hi John - is there any chance this crossover routine can make its way into
one of the releases. I am seeing lots of issues in CLP/CBC at trunk (see
other emails) - so I switched back to releases Osi (0.100.2), Clp (1.10.2),
Cbc (2.3.2). The releases seem better - but don't have the crossover
function.

Thanks,
Matt


Matt,
>
> I have added some code to OsiClp. A very simple example of its use is in
> Cbc/examples!! I know that is the wrong place, but there is nowhere to put a
> OsiClp specific example.
>
> John
>
> [image: Inactive hide details for Matthew Galati ---10/07/2009 12:24:32
> AM---I looked at the code where you do crossover - and there is]Matthew
> Galati ---10/07/2009 12:24:32 AM---I looked at the code where you do
> crossover - and there is quite a bit of logic there. And it also s
>
>
> From:
> Matthew Galati <magh at lehigh.edu>
> To:
> John J Forrest/Watson/IBM at IBMUS, Matthew Galati <matthew.galati at sas.com>
> Date:
> 10/07/2009 12:24 AM
>
> Subject:
> Re: Fwd: crossover step
> ------------------------------
>
>
>
> I looked at the code where you do crossover - and there is quite a bit of
> logic there. And it also seems tied into the barrier data structs. Is it
> possible for you to create a function in OsiClp that takes in a current
> solution and forms a basis from that? I.e., basically a warm start for
> OsiClp but in the form of a primal solution, not a basis.
>
>
>
>    Matt,
>
>    Code is in ClpSimplex:initialSolve which confusingly is in ClpSolve.cpp
>    as ClpSolve controls options. Round about 2032 is where code returns from
>    barrier.
>
>    By the way, many bugs have been fixed (not exactly by mistake) in
>    trunk. Do you know if things work better with trunk? I will try and get a
>    new stable by early December.
>
>    John
>
>    [image: Inactive hide details for Matthew Galati ---10/05/2009 05:29:44
>    PM---Hi John - do you think this is possible using CLP? I assum]Matthew
>    Galati ---10/05/2009 05:29:44 PM---Hi John - do you think this is possible
>    using CLP? I assume you do something like this for interior
>
>
>
> From:
> Matthew Galati <*magh at lehigh.edu* <magh at lehigh.edu>>
> To:
> John J Forrest/Watson/IBM at IBMUS
> Date:
> 10/05/2009 05:29 PM
> Subject:
> Fwd: crossover step
>
>    ------------------------------
>
>
>
>
>    Hi John - do you think this is possible using CLP? I assume you do
>    something like this for interior point to crossover to simplex? Can you
>    point me to where in your code you do this and any suggestion on how I can
>    extract that to do it from OSI layer?
>
>    Thanks in advance,
>    Matt
>
>    ---------- Forwarded message ----------
>    From: *Matthew Galati* <*magh at lehigh.edu* <magh at lehigh.edu>>
>    Date: Mon, Sep 7, 2009 at 11:51 PM
>    Subject: crossover step
>    To: *clp at list.coin-or.org* <clp at list.coin-or.org>
>
>
>    After interior completes - it can/does(?) do a crossover step and then
>    completes with simplex. Can someone point me to the piece of code that does
>    the crossover step in CLP?
>
>    I am assuming there is no Osi interface for this (crossover)? It would
>    be nice if there was.
>
>    Thanks,
>    Matt
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/osi/attachments/20091026/94991dc4/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ecblank.gif
Type: image/gif
Size: 45 bytes
Desc: not available
URL: <http://list.coin-or.org/pipermail/osi/attachments/20091026/94991dc4/attachment.gif>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: graycol.gif
Type: image/gif
Size: 105 bytes
Desc: not available
URL: <http://list.coin-or.org/pipermail/osi/attachments/20091026/94991dc4/attachment-0001.gif>