No solution returned although Cbc claim optimality

[andreaslundell]

I have encountered a problem when solving the following MILP problem in Cbc:

Minimize
 10.5 i1 - 7.5 i2 - 3.5 i3 + 2.5 i4 - 1.5 i5 + 10 i6 + 0.5 x7 + 0.25 x8 + 1.5 x9 + 0.5 x10 + 0.5 x11
Subject To
 e1: 6 i1 + 3 i2 + 3 i3 + 2 i4 + i5 <= 6.5
 e2: 10 i1 + 10 i3 + i6 <= 20
 e3: 1.13238 i3 - x9 <= 0.32057
 e4: 1.23607 i5 - x11 <= 0.38197
 e5: 1.23607 i2 - x8 <= 0.38197
 e6: 1.5982 i3 - x9 <= 0.63856
 e7: 1.85942 i5 - x11 <= 0.86436
 e8: 1.85942 i2 - x8 <= 0.86436
 e9: 1.99507 i2 - x8 <= 0.99507
 e10: 1.96005 i3 - x9 <= 0.96045
 e11: 10.5 i1 - 7.5 i2 - 3.5 i3 + 2.5 i4 - 1.5 i5 + 10 i6 + 0.5 x7 + 0.25 x8 + 1.5 x9 + 0.5 x10 + 0.5 x11 <= -9.25
Bounds
 0 <= i1 <= 1
 0 <= i2 <= 1
 0 <= i3 <= 1
 0 <= i4 <= 1
 0 <= i5 <= 1
 0 <= i6 <= 20
 0 <= x7 <= 1
 0 <= x8 <= 1
 0 <= x9 <= 1
 0 <= x10 <= 1
 0 <= x11 <= 1
Integers
 i1 i2 i3 i4 i5 i6
End

When solving it with the trunk version of Cbc, I get the following output:

Welcome to the CBC MILP Solver
Version: Trunk (unstable)
Build Date: Dec  1 2019

command line - /opt/cbc-master/bin/cbc test.lp (default strategy 1)
Continuous objective value is -10.1292 - 0.00 seconds
Cgl0004I processed model has 0 rows, 0 columns (0 integer (0 of which binary)) and 0 elements
Cbc3007W No integer variables - nothing to do
Cuts at root node changed objective from -9.2506 to -1.79769e+308
Probing was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
Gomory was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
Knapsack was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
Clique was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
MixedIntegerRounding2 was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
FlowCover was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
TwoMirCuts was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
ZeroHalf was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)

Result - Optimal solution found

Objective value:                -9.25060000
Enumerated nodes:               0
Total iterations:               3
Time (CPU seconds):             0.00
Time (Wallclock seconds):       0.00

Total time (CPU seconds):       0.00   (Wallclock seconds):       0.00

So far so good... However, if the solution is written to a file using (cbc test.lp solu sol.txt) the correct solution is not returned:

sol.txt:

Status unknown - objective value 0.00000000
      0 i1                     0                       0
      1 i2                     0                       0
      2 i3                     0                       0
      3 i4                     0                       0
      4 i5                     0                       0
      5 i6                     0                       0
      6 x7                     0                       0
      7 x8                     0                       0
      8 x9                     0                       0
      9 x10                    0                       0
     10 x11                    0                       0

Solving the same problem with Gurobi (and CPLEX) gives the correct solution:

Gurobi Optimizer version 8.1.0 build v8.1.0rc1 (linux64)
Copyright (c) 2018, Gurobi Optimization, LLC

Read LP format model from file mod.lp
Reading time = 0.00 seconds
: 11 rows, 11 columns, 35 nonzeros
Optimize a model with 11 rows, 11 columns and 35 nonzeros
Variable types: 5 continuous, 6 integer (0 binary)
Coefficient statistics:
  Matrix range     [2e-01, 1e+01]
  Objective range  [2e-01, 1e+01]
  Bounds range     [1e+00, 2e+01]
  RHS range        [3e-01, 2e+01]
Presolve removed 11 rows and 11 columns
Presolve time: 0.00s
Presolve: All rows and columns removed

Explored 0 nodes (0 simplex iterations) in 0.00 seconds
Thread count was 1 (of 6 available processors)

Solution count 1: -9.2506 

Optimal solution found (tolerance 1.00e-04)
Best objective -9.250600000000e+00, best bound -9.250600000000e+00, gap 0.0000%

Solution:

# Objective value = -9.2506
i1 0
i2 1
i3 1
i4 0
i5 0
i6 0
x7 0
x8 9.9999999999999989e-01
x9 0.9996
x10 0
x11 0

[tkralphs]

Looks like the solution was found in pre-processing, but because the pre-processed model was completely empty, something went wrong in getting the solution out. We'll take a look.

Does the same happen with the latest release? Keep in mind that trunk/master version of Cbc is under active development and some bugs are to be expected.

[jjhforrest]

Should be cbc test.lp -solve -solu sol.txt If no parameters are given then does solve - otherwise does as told. John Forrest

…

On 01/12/2019 16:46, Andreas Lundell wrote: I have encountered a problem when solving the following MILP problem in Cbc: |Minimize 10.5 i1 - 7.5 i2 - 3.5 i3 + 2.5 i4 - 1.5 i5 + 10 i6 + 0.5 x7 + 0.25 x8 + 1.5 x9 + 0.5 x10 + 0.5 x11 Subject To e1: 6 i1 + 3 i2 + 3 i3 + 2 i4 + i5 <= 6.5 e2: 10 i1 + 10 i3 + i6 <= 20 e3: 1.13238 i3 - x9 <= 0.32057 e4: 1.23607 i5 - x11 <= 0.38197 e5: 1.23607 i2 - x8 <= 0.38197 e6: 1.5982 i3 - x9 <= 0.63856 e7: 1.85942 i5 - x11 <= 0.86436 e8: 1.85942 i2 - x8 <= 0.86436 e9: 1.99507 i2 - x8 <= 0.99507 e10: 1.96005 i3 - x9 <= 0.96045 e11: 10.5 i1 - 7.5 i2 - 3.5 i3 + 2.5 i4 - 1.5 i5 + 10 i6 + 0.5 x7 + 0.25 x8 + 1.5 x9 + 0.5 x10 + 0.5 x11 <= -9.25 Bounds 0 <= i1 <= 1 0 <= i2 <= 1 0 <= i3 <= 1 0 <= i4 <= 1 0 <= i5 <= 1 0 <= i6 <= 20 0 <= x7 <= 1 0 <= x8 <= 1 0 <= x9 <= 1 0 <= x10 <= 1 0 <= x11 <= 1 Integers i1 i2 i3 i4 i5 i6 End | When solving it with the trunk version of Cbc, I get the following output: |Welcome to the CBC MILP Solver Version: Trunk (unstable) Build Date: Dec 1 2019 command line - /opt/cbc-master/bin/cbc test.lp (default strategy 1) Continuous objective value is -10.1292 - 0.00 seconds Cgl0004I processed model has 0 rows, 0 columns (0 integer (0 of which binary)) and 0 elements Cbc3007W No integer variables - nothing to do Cuts at root node changed objective from -9.2506 to -1.79769e+308 Probing was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) Gomory was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) Knapsack was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) Clique was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) MixedIntegerRounding2 was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) FlowCover was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) TwoMirCuts was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) ZeroHalf was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) Result - Optimal solution found Objective value: -9.25060000 Enumerated nodes: 0 Total iterations: 3 Time (CPU seconds): 0.00 Time (Wallclock seconds): 0.00 Total time (CPU seconds): 0.00 (Wallclock seconds): 0.00 | So far so good... However, if the solution is written to a file using (|cbc test.lp solu sol.txt|) the correct solution is not returned: sol.txt: |Status unknown - objective value 0.00000000 0 i1 0 0 1 i2 0 0 2 i3 0 0 3 i4 0 0 4 i5 0 0 5 i6 0 0 6 x7 0 0 7 x8 0 0 8 x9 0 0 9 x10 0 0 10 x11 0 0 | Solving the same problem with Gurobi (and CPLEX) gives the correct solution: |Gurobi Optimizer version 8.1.0 build v8.1.0rc1 (linux64) Copyright (c) 2018, Gurobi Optimization, LLC Read LP format model from file mod.lp Reading time = 0.00 seconds : 11 rows, 11 columns, 35 nonzeros Optimize a model with 11 rows, 11 columns and 35 nonzeros Variable types: 5 continuous, 6 integer (0 binary) Coefficient statistics: Matrix range [2e-01, 1e+01] Objective range [2e-01, 1e+01] Bounds range [1e+00, 2e+01] RHS range [3e-01, 2e+01] Presolve removed 11 rows and 11 columns Presolve time: 0.00s Presolve: All rows and columns removed Explored 0 nodes (0 simplex iterations) in 0.00 seconds Thread count was 1 (of 6 available processors) Solution count 1: -9.2506 Optimal solution found (tolerance 1.00e-04) Best objective -9.250600000000e+00, best bound -9.250600000000e+00, gap 0.0000% | Solution: |# Objective value = -9.2506 i1 0 i2 1 i3 1 i4 0 i5 0 i6 0 x7 0 x8 9.9999999999999989e-01 x9 0.9996 x10 0 x11 0 | — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#273?email_source=notifications&email_token=ABWJYHFKLN2OFEJ5QFBT6ULQWPS6XA5CNFSM4JTMR3MKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4H5CX5CA>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABWJYHCAJI3VHEOO2IYZAP3QWPS6XANCNFSM4JTMR3MA>.

[andreaslundell]

Using -solu did indeed print the correct solution. I was following the instructions at https://projects.coin-or.org/CoinBinary/export/1059/OptimizationSuite/trunk/Installer/files/doc/cbcCommandLine.pdf, which had just solu. This does not however explain the problems I have solving this problem using the C++ interface, but I will investigate that further...