[Cbc] Bigger paradoxical case for extraVariables option
acw at ascent.com
acw at ascent.com
Mon Jun 6 10:28:03 EDT 2016
This is a sequel to a problem I posted about back in April.
I have a 3.4 Mb MPS file that has the following behavior:
cbc -import extrav-test.mps -extrav 10 -solve -quit
takes 9723 CPU seconds to solve, while
cbc -import extrav-test.mps -solve -quit
takes 10850 CPU seconds. The difference is only about 10%, but it's in the
wrong direction -- that is, the -extraVariables trick makes it solve
slower. In April I presented a similar case, but Dr. Forrest pointed out
on that occasion that the solver spent most of its time on heuristics, and
the problem was in fact solved by heuristic; apparently introducing the
extra variables slows down the heuristic phase.
But the new problem spends most of its time branching, and -extrav still
slows it down, though not as dramatically as the earlier problem. I can't
send the MPS file to the list because it's too big, but I will be happy to
share it with anybody who is willing to take a look.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20160606/46158988/attachment.html>
More information about the Cbc
mailing list