[Cbc] Baffling (resend)
John Forrest
john.forrest at fastercoin.com
Wed Mar 5 15:10:26 EST 2014
David,
Could be something to do with restarting search. It looks as if Cbc did
100 nodes and saw it could fix variables (because of getting a solution)
to make problem smaller. So it solves that smaller problem.
a) might explain times - maybe it took 900 seconds to do those first
nodes???
b) may be a bug with max time and restarting (may also explain
preprocessing problem)
What happens if you add -strategy 0 - that should switch off restart?
If you want, send the mps file to me - I presume it is only a few
megabytes compressed.
John Forrest
On 05/03/14 18:51, David Einstein wrote:
> (attaching large mps files is not allowed, so I'm resending)
>
> I am calling libCbcSolver as follows ( cbc 2.8 windows 7 64 bit)
>
> CbcModel cbc(osi);
> CbcMain0(cbc);
> cbc.setDblParam(CbcModel::CbcMaximumSeconds, 1800);
> const char *argv2[] = {"schedulemip", "-preprocess", "off",
> "-solve", "-quit"};
> CbcMain1(5, argv2, cbc);
>
>
> and the tail end of the output looks like
> ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Cbc0010I After 10800 nodes, 1166 on tree, 169460.67 best solution,
> best possible 157572.58 (804.68 seconds)
> Cbc0010I After 10900 nodes, 1214 on tree, 169460.67 best solution,
> best possible 157572.58 (810.69 seconds)
> Cbc0010I After 11000 nodes, 1264 on tree, 169460.67 best solution,
> best possible 157572.58 (815.13 seconds)
> Cbc0010I After 11100 nodes, 1311 on tree, 169460.67 best solution,
> best possible 158109.85 (835.74 seconds)
> Cbc0010I After 11200 nodes, 1361 on tree, 169460.67 best solution,
> best possible 158345.27 (853.65 seconds)
> Cbc0010I After 11300 nodes, 1411 on tree, 169460.67 best solution,
> best possible 158404.02 (869.75 seconds)
> Cbc0010I After 11400 nodes, 1461 on tree, 169460.67 best solution,
> best possible 158464.46 (886.81 seconds)
> Cbc0020I Exiting on maximum time
> Cbc0005I Partial search - best objective 169460.67 (best possible
> 158464.46), took 1871577 iterations and 11482 nodes (900.43 seconds)
> Cbc0032I Strong branching done 8334 times (76189 iterations),
> fathomed 9 nodes and fixed 1 variables
> Cbc0035I Maximum depth 124, 223498 variables fixed on reduced cost
> Cbc0038I Probing was tried 1579 times and created 82296 cuts of
> which 1 were active after adding rounds of cuts (11.281 seconds)
> Cbc0038I Gomory was tried 7626 times and created 8129 cuts of
> which 2464 were active after adding rounds of cuts (144.295 seconds)
> Cbc0038I Knapsack was tried 2 times and created 0 cuts of which 0
> were active after adding rounds of cuts (0.016 seconds)
> Cbc0038I Clique was tried 2 times and created 0 cuts of which 0
> were active after adding rounds of cuts (0.000 seconds)
> Cbc0038I MixedIntegerRounding2 was tried 2 times and created 0
> cuts of which 0 were active after adding rounds of cuts (0.031
> seconds)
> Cbc0038I FlowCover was tried 2 times and created 0 cuts of which 0
> were active after adding rounds of cuts (0.000 seconds)
> Cbc0038I TwoMirCuts was tried 1579 times and created 10 cuts of
> which 1 were active after adding rounds of cuts (3.860 seconds)
> Cbc0038I Stored from first was tried 7626 times and created 350
> cuts of which 179 were active after adding rounds of cuts (0.000
> seconds)
> Cbc0001I Search completed - best objective 184071.9400000001, took
> 1886804 iterations and 11532 nodes (900.65 seconds)
> Cbc0032I Strong branching done 428 times (5741 iterations),
> fathomed 0 nodes and fixed 0 variables
> Cbc0035I Maximum depth 34, 0 variables fixed on reduced cost
> Cuts at root node changed objective from 156922 to 157294
> Probing was tried 3 times and created 0 cuts of which 0 were
> active after adding rounds of cuts (0.032 seconds)
> Gomory was tried 24 times and created 61 cuts of which 28 were
> active after adding rounds of cuts (1.650 seconds)
> Knapsack was tried 3 times and created 0 cuts of which 0 were
> active after adding rounds of cuts (0.016 seconds)
> Clique was tried 3 times and created 0 cuts of which 0 were active
> after adding rounds of cuts (0.000 seconds)
> MixedIntegerRounding2 was tried 3 times and created 0 cuts of
> which 0 were active after adding rounds of cuts (0.031 seconds)
> FlowCover was tried 1 times and created 0 cuts of which 0 were
> active after adding rounds of cuts (0.000 seconds)
> TwoMirCuts was tried 3 times and created 5 cuts of which 0 were
> active after adding rounds of cuts (0.624 seconds)
>
> Result - Optimal solution found
>
> Objective value: 184071.94000000
> Enumerated nodes: 11532
> Total iterations: 1886804
> Time (CPU seconds): 901.01
> Time (Wallclock seconds): 901.01
>
> Total time (CPU seconds): 901.50 (Wallclock seconds):
> 901.49
>
> ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Things that baffle me:
> 1) It is timing out after 900 seconds instead of 1800
> 2) It is reporting 184071.94 as optimal when it has found a solution
> of 169460.67
>
> I am running with preprocesssing off because I get the following with
> it on
>
> Coin0505I Presolved problem not optimal, resolve after postsolve
> Coin0505I Presolved problem not optimal, resolve after postsolve
> Cgl0013I Postprocessed model is infeasible - possible tolerance
> issue - try without preprocessing
> 1 relaxed row infeasibilities - summing to 1
> 1 relaxed row infeasibilities - summing to 1
> 1 relaxed row infeasibilities - summing to 1
>
> Result - Stopped on time limit
>
> Objective value: 173347.10500000
> Lower bound: 159327.768
> Gap: 0.09
> Enumerated nodes: 23597
> Total iterations: 4755507
> Time (CPU seconds): 1800.63
>
>
>
> Am I doing something wrong?
>
>
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/cbc
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20140305/79e62457/attachment-0001.html>
More information about the Cbc
mailing list