<html><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /></head><body style='font-size: 10pt; font-family: Verdana,Geneva,sans-serif'>
<p>Dear Ambros,</p>
<p>That's was exactly the problem. The problem we implemented in SCIP was different to the one in our notes.</p>
<p>Attached you have the problem using the correct bounds. Now everything makes sense. It's time to use the lazy bounds for bigger problems which are not solved at the root node.</p>
<p>Thank you so much!</p>
<p>Best.</p>
<p>Diego.</p>
<div> </div>
<p><br /></p>
<p>El 26/04/2018 04:27, Ambros Gleixner escribió:</p>
<blockquote type="cite" style="padding: 0 0.4em; border-left: #1010ff 2px solid; margin: 0"><!-- html ignored --><!-- head ignored --><!-- meta ignored -->
<div class="pre" style="margin: 0; padding: 0; font-family: monospace"><span style="white-space: nowrap;">Dear Diego,</span><br /> <br /> I am a bit suspicious about the initial dual solution with value -10. The log shows that SCIP finds a -24 primal solution as solution of the initial LP solve (symbol "*" on the left). This looks like SoPlex returns a dual solution with value -24.<br /> <br /> When you calculate your dual solution, do you take into account that there may be nonzero dual multipliers for the upper bounds of your variables? If not, you can avoid this by declaring them to be lazy, see FAQ<br /> <br /> <span style="white-space: nowrap;"> <a href="http://scip.zib.de/doc-5.0.1/html/FAQ.php#whatarelazybounds" target="_blank" rel="noreferrer">http://scip.zib.de/doc-5.0.1/html/FAQ.php#whatarelazybounds</a></span><br /> <br /> and the use of SCIPchgVarUbLazy() in examples/Binpacking/src/pricer_binpacking.c.<br /> <br /> <span style="white-space: nowrap;">Best,</span><br /> Ambros<br /> <br /> <br /> <span style="white-space: nowrap;">Am 25.04.2018 um 14:05 schrieb dponce@us.es:</span>
<blockquote type="cite" style="padding: 0 0.4em; border-left: #1010ff 2px solid; margin: 0"><span style="white-space: nowrap;">Dear Ambros,</span><br /> <br /> I thought exactly the same. It would be a natural explanation for the behavior. But you can see in the attached output the same oscilation, setting the flag removable to FALSE in the function SCIPcreateVar().<br /> <br /> <span style="white-space: nowrap;">Best.</span><br /> <br /> <span style="white-space: nowrap;">Diego.</span><br /> <br /> <br /> <span style="white-space: nowrap;">El 25/04/2018 10:41, Ambros Gleixner escribió:</span><br /> <br />
<blockquote type="cite" style="padding: 0 0.4em; border-left: #1010ff 2px solid; margin: 0"><span style="white-space: nowrap;">Dear Diego,</span><br /> <br /> <span style="white-space: nowrap;">Sorry that we have been slow in answering. It seems like a tricky question.</span><br /> <br /> <br /> <span style="white-space: nowrap;">Am 17.04.2018 um 17:20 schrieb dponce@us.es:</span>
<blockquote type="cite" style="padding: 0 0.4em; border-left: #1010ff 2px solid; margin: 0"><span style="white-space: nowrap;">Dear SCIP,</span><br /> <br /> Developing a B&P project, we have some questions about the behavior we have among the different iterations of the pricer.<br /> <br /> Attached you have the output of a set partitioning formulation for which eventually the optimal solution is reached. But still I see a couple of things I don't understand:<br /> <br /> <span style="white-space: nowrap;"> * Line 109. Why the first solution of the dual problem is not -24.0?</span><br /> <span style="white-space: nowrap;"> With the initial variables the dual problem should have constraints</span><br /> <span style="white-space: nowrap;"> to avoid solution -10.0.</span><br /> <span style="white-space: nowrap;"> * Line 279. If previously the solution was -114.0, why suddenly goes</span><br /> <span style="white-space: nowrap;"> back to -112.0?</span></blockquote>
<br /> It seems that many LP columns have been removed from the LP, since the value in the output column labeled "cols" drops from 62 to 29. I am not an expert user of branch-and-price in SCIP, does somebody have a suggestion how to analyze/avoid the behavior?<br /> <br /> <br />
<blockquote type="cite" style="padding: 0 0.4em; border-left: #1010ff 2px solid; margin: 0"><span style="white-space: nowrap;">Regarding this previous question</span><br /> <br /> <span style="white-space: nowrap;"><a href="http://listserv.zib.de/pipermail/scip/2017-July/003156.html" target="_blank" rel="noreferrer">http://listserv.zib.de/pipermail/scip/2017-July/003156.html</a></span><br /> <br /> <span style="white-space: nowrap;">and the binpacking example, we have disabled the preprocessing.</span></blockquote>
<br /> During pricing you can always query the dual solution for the transformed problem, so there is no need to disable presolving. This is only necessary if you want to obtain the dual solution for an LP that you solve with SCIP "out of the box".<br /> <br /> <br />
<blockquote type="cite" style="padding: 0 0.4em; border-left: #1010ff 2px solid; margin: 0"><span style="white-space: nowrap;">Is it disabled in the VRP example?</span></blockquote>
<br /> <span style="white-space: nowrap;">I don't think so.</span><br /> <br /> <span style="white-space: nowrap;">Best,</span><br /> Ambros<br /> <br /> <br /> <br />
<blockquote type="cite" style="padding: 0 0.4em; border-left: #1010ff 2px solid; margin: 0">Dual solution has been handily calculated from the dual multiplier. Are there some SCIP function to make this for the current solution? And for primal? Anyway, we trust on our calculation and this is a way to check the correctness of the dual multiplier values.<br /> <br /> We are debugging our code, but right now, we don't have more ideas. We have thought in DELAY and removable option in the variables as an explanation of this behavior, but nothing.<br /> <br /> <span style="white-space: nowrap;">Thanks in advance for the help.</span><br /> <br /> <span style="white-space: nowrap;">Best.</span><br /> <br /> <span style="white-space: nowrap;">Diego Ponce.</span><br /> <br /> <br /> <br /> <br /> _______________________________________________<br /> <span style="white-space: nowrap;">Scip mailing list</span><br /> <span style="white-space: nowrap;"><a href="mailto:Scip@zib.de" rel="noreferrer">Scip@zib.de</a> <mailto:<a href="mailto:Scip@zib.de" rel="noreferrer">Scip@zib.de</a>></span><br /> <span style="white-space: nowrap;"><a href="https://listserv.zib.de/mailman/listinfo/scip" target="_blank" rel="noreferrer">https://listserv.zib.de/mailman/listinfo/scip</a></span></blockquote>
</blockquote>
</blockquote>
</div>
</blockquote>
</body></html>