<div dir="ltr"><div>Just a precision.<br><br>In my model, there is a valid lower bound on the master problem continuous relaxation value.<br></div><div>I stop column generation when a restricted master problem continuous relaxation value is equal to this bound,<br>
</div><div>even if the lagrangian bound is still oscillating.<br></div><div>For instances of practical interest, this bound is always met quickly.<br></div><div>So, as my initialization is good, no columns are generated for a long time in a B&P execution.<br>
</div><div>Only LP solving (without CG) and branching ...<br></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">2014-02-21 9:55 GMT+01:00 Xavier Schepler <span dir="ltr"><<a href="mailto:xavier.schepler@gmail.com" target="_blank">xavier.schepler@gmail.com</a>></span>:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">
<div text="#000000" bgcolor="#FFFFFF">
<div>Hi professor Lübbecke,<br>
<br>The B&P is working fine on small instances.<br></div><div>For some bigger instances, most of the computing time is spent in solving linear relaxations of restricted master problems.<br></div><div class=""><br>
<blockquote type="cite">
<div dir="ltr">
<div>
<div> Can you plot the objective function values (over
iterations) of the restricted master problem (to check
whether you have strong tailing off)?<br></div>
</div>
</div>
</blockquote>
<blockquote type="cite">
<div dir="ltr">
<div>
<div>Did you check dual variable development (you can plot
this as well) to check whether you have unstable duals?<br>
</div>
</div>
</div>
</blockquote>
</div></div><div text="#000000" bgcolor="#FFFFFF">Tailing off is a real problem, since an optimal value to the linear relaxation of a master problem (v_mp) is attained in a few column generation iterations,<br>but, the lagrangian bound is very far away, and needs many more iterations to meet v_mp . <br>
</div><div text="#000000" bgcolor="#FFFFFF"><div class="">
<br>
<blockquote type="cite">
<div dir="ltr">
<div>
<div> However, you could tell us about your branching rule.
Branch on master vars (uah...) or on other information (like
"original" vars)? Ryan-Foster?</div>
</div>
</div>
</blockquote></div>
Branching is done on original variables, and branching constraints are added to the structural constraints,<br>as it is described in Gerald Gamrath thesis.<div class=""><br>
<br>
<blockquote type="cite">
<div dir="ltr">
<div>
<div> Give us a little more detail. [BTW, you fix the *local*
UBs of the vars, right?]<br>
<br>
</div>
</div>
</div>
</blockquote></div>
Well, I fix local upper bounds.<div class=""><br>
<br>
<blockquote type="cite">
<div dir="ltr">
<div> Out of
curiosity: what is "time limit"</div>
</div>
</blockquote></div>
I don't know exactly what the time limit is, but SCIP outputs :<br>
LP solver hit time limit.<div class=""><br>
<br>
<blockquote type="cite">
<div dir="ltr">
<div> and "a very long time?"<br>
</div>
</div>
</blockquote></div>
Maybe several minutes, which seems quite long for LP
re-optimization.<br><br></div><div text="#000000" bgcolor="#FFFFFF">I would like SoPlex to display information ... <br></div><div text="#000000" bgcolor="#FFFFFF">
<br></div><div text="#000000" bgcolor="#FFFFFF">Best regards,<br><br>Xavier<br>
</div>
</div>
</blockquote></div><br></div>