[Cbc-tickets] [COIN-OR Branch-and-Cut MIP Solver] #58: Cbc 2.2 reports model integer-infeasible, actually feasible
COIN-OR Branch-and-Cut MIP Solver
coin-trac at coin-or.org
Fri Oct 3 06:05:31 EDT 2008
#58: Cbc 2.2 reports model integer-infeasible, actually feasible
-------------------------+--------------------------------------------------
Reporter: nowozin | Owner: somebody
Type: defect | Status: new
Priority: major | Milestone:
Component: component1 | Version:
Resolution: | Keywords:
-------------------------+--------------------------------------------------
Comment (by nowozin):
This is most likely a bug in Cbc 2.2 or Cgl.
The MPS file available at http://www.nowozin.net/test.zip produces a
integer-feasible solution after 25,000 nodes with GLPK 4.27:
{{{
glpsol --mps test.mps
lpx_read_mps: reading problem data from `test.mps'...
lpx_read_mps: problem name not specified
lpx_read_mps: 2690 rows, 4032 columns, 270144 non-zeros
lpx_read_mps: 4032 integer columns, 2016 of which are binary
lpx_read_mps: 141138 cards were read
glp_simplex: original LP has 2690 rows, 4032 columns, 270144 non-zeros
glp_simplex: presolved LP has 1681 rows, 3024 columns, 135072 non-zeros
lpx_adv_basis: size of triangular part = 1681
0: objval = 0.000000000e+00 infeas = 1.000000000e+00 (0)
200: objval = 0.000000000e+00 infeas = 1.000000000e+00 (0)
400: objval = 4.968000000e+04 infeas = 4.858684576e-01 (0)
600: objval = 6.461333333e+04 infeas = 3.389731870e-01 (0)
800: objval = 8.531630966e+04 infeas = 1.502789781e-01 (0)
1000: objval = 9.835184102e+04 infeas = 6.772932409e-02 (0)
1200: objval = 1.007848246e+05 infeas = 4.777066272e-02 (0)
1400: objval = 1.036354949e+05 infeas = 3.453695403e-02 (0)
1600: objval = 1.058332710e+05 infeas = 1.789188085e-02 (0)
1800: objval = 1.050143791e+05 infeas = 9.124520507e-03 (0)
2000: objval = 1.053971785e+05 infeas = 4.370541393e-03 (0)
2200: objval = 1.060234909e+05 infeas = 2.644864612e-03 (0)
2400: objval = 1.061356559e+05 infeas = 5.741590605e-04 (0)
2551: objval = 1.059470340e+05 infeas = 2.332458220e-18 (0)
* 2551: objval = 1.059470340e+05 infeas = 2.000479442e-13 (0)
* 2600: objval = 1.051352657e+05 infeas = 2.574980160e-18 (0)
* 2800: objval = 1.025373690e+05 infeas = 4.529709940e-16 (0)
* 3000: objval = 1.006550013e+05 infeas = 1.366278924e-15 (0)
* 3200: objval = 9.961699583e+04 infeas = 3.252606517e-19 (0)
* 3400: objval = 9.933354845e+04 infeas = 5.684402872e-14 (0)
* 3600: objval = 9.920301384e+04 infeas = 5.487061903e-18 (0)
* 3800: objval = 9.912210407e+04 infeas = 4.380854403e-18 (0)
* 3937: objval = 9.910482706e+04 infeas = 4.619415961e-12 (0)
OPTIMAL SOLUTION FOUND
Integer optimization begins...
+ 3937: mip = not found yet >= -inf (1; 0)
+ 3942: mip = not found yet >= 9.910482706e+04 (4; 0)
+ 3945: mip = not found yet >= 9.910482706e+04 (7; 0)
+ 3947: mip = not found yet >= 9.910482706e+04 (9; 0)
...
+ 25400: mip = not found yet >= 9.910482706e+04 (801; 0)
+ 25466: mip = not found yet >= 9.910482706e+04 (821; 0)
+ 25498: >>>>> 1.202200000e+05 >= 9.910482706e+04 17.6% (834; 0)
+ 25676: mip = 1.202200000e+05 >= 9.910482706e+04 17.6% (828; 17)
}}}
Whereas Cbc 2.2 run on the same file produces:
{{{
Coin Cbc and Clp Solver version 2.20.00, build Sep 20 2008
command line - /opt/coin-cbc-2.2.0/bin/cbc test.mps
At line 8 NAME
At line 9 ROWS
At line 2700 COLUMNS
At line 137775 RHS
At line 138113 BOUNDS
At line 141138 ENDATA
Problem no_name has 2689 rows, 4032 columns and 268128 elements
Coin0008I no_name read with 0 errors
Continuous objective value is 145980 - 0.14 seconds
Perturbing problem by 0.001 % of 478.647 - largest nonzero change
0.00146329 (% 0.000305713) - largest zero change 0.000244159
0 Obj 145980 Primal inf 3349.78 (576)
Primal infeasible - objective value 153640
Cgl0000I Cut generators found to be infeasible!
Pre-processing says infeasible or unbounded
Total time 0.70
}}}
--
Ticket URL: <https://projects.coin-or.org/Cbc/ticket/58#comment:3>
COIN-OR Branch-and-Cut MIP Solver <http://projects.coin-or.org/Cbc>
An LP-based branch-and-cut MIP solver.
More information about the Cbc-tickets
mailing list