Trying to use CBC within google OR Mixed Integer Programming framework to solve processed model has 1029 rows, 7240 columns (7196 integer (7196 of which binary)) and 249956 elements

[Charles Rosa]

To all,

I am trying to use CBC within google OR Mixed Integer Programming framework to solve a Mixed Integer Programming model.  The processed model has 1029 rows, 7240 columns (7196 integer (7196 of which binary)) and 249956 elements.

The CBC solver returns almost immediately with the messages:

[16:05:39] src/linear_solver/linear_solver.cc:1215: No solution exists. MPSolverInterface::result_status_ = MPSOLVER_NOT_SOLVED

Exiting on maximum time

Partial search - best objective 1e+50 (best possible 10729.942), took 0 iterations and 0 nodes (0.72 seconds)

Maximum depth 0, 0 variables fixed on reduced cost

[16:06:00] src/linear_solver/linear_solver.cc:1215: No solution exists. MPSolverInterface::result_status_ = MPSOLVER_NOT_SOLVED

2017-07-24 16:06:05,700 [main] DEBUG AggregationModelCplexSolver  - Optimizing iteration 1 done

2017-07-24 16:06:05,700 [main] DEBUG AggregationModelCplexSolver  - Optimizing iteration 1 done

2017-07-24 16:06:11,541 [main] DEBUG AggregationModelCplexSolver  - Total time 67.0

2017-07-24 16:06:11,541 [main] DEBUG AggregationModelCplexSolver  - Total time 67.0

2017-07-24 16:06:12,621 [main] DEBUG AggregationModelCplexSolver  -  solution - 0.0 bound 0.0

2017-07-24 16:06:12,621 [main] DEBUG AggregationModelCplexSolver  -  solution - 0.0 bound 0.0

2017-07-24 16:06:17,118 [main] DEBUG AggregationModelCplexSolver  -  abs gap 0.0; gap NaN

2017-07-24 16:06:17,118 [main] DEBUG AggregationModelCplexSolver  -  abs gap 0.0; gap NaN

2017-07-24 16:06:17,525 [main] DEBUG AggregationModelCplexSolver  -  status NOT_SOLVED cutOff value 0.0

2017-07-24 16:06:17,525 [main] DEBUG AggregationModelCplexSolver  -  status NOT_SOLVED cutOff value 0.0

2017-07-24 16:07:26,313 [main] ERROR main.LaunchPoint  - failed!

However, if I solve the exact same model with the commerical solver CPLEX, I get a solution.

I've solved smaller MIPs (toy models basically) using CBC and google's OR Mixed Integer Programming framework.

For the smaller model I get the following output:

processed model has 13 rows, 21 columns (16 integer (16 of which binary)) and 68 elements

Full problem 14 rows 96 columns, reduced to 13 rows 21 columns

The LP relaxation is infeasible or too expensive

At root node, 0 cuts changed objective from 65.248498 to 72.577697 in 1 passes

Cut generator 0 (Probing) - 4 row cuts average 2.0 elements, 2 column cuts (2 active)

Search completed - best objective 1e+50, took 0 iterations and 0 nodes (0.02 seconds)

Maximum depth 0, 0 variables fixed on reduced cost

Mini branch and bound did not improve solution (0.02 seconds)

After 0.02 seconds - Feasibility pump exiting with objective of 71.2502 - took 0.01 seconds

Integer solution of 71.2502 found by feasibility pump after 0 iterations and 0 nodes (0.02 seconds)

processed model has 5 rows, 6 columns (3 integer (3 of which binary)) and 16 elements

Full problem 14 rows 96 columns, reduced to 5 rows 6 columns

The LP relaxation is infeasible or too expensive

At root node, 0 cuts changed objective from 65.248498 to 65.248498 in 1 passes

Cut generator 0 (Probing) - 1 row cuts average 0.0 elements, 1 column cuts (1 active)

Search completed - best objective 1e+50, took 0 iterations and 0 nodes (0.02 seconds)

Maximum depth 0, 0 variables fixed on reduced cost

Does anyone know how to get CBC to solve the modest sized (1029 rows, 7240 columns (7196 integer (7196 of which binary)) and 249956 elements) problem first described above?

I can't even figure out, currently, how to change the default parameters of the solver through the google OR framework (or even to figure out what the default values are).

Thank you for whatever thoughts/advice you can give.

Best regards, Charlie

[Charles Rosa]

I forgot to mention.....

I'm using the Java interface into the google OR Mixed Integer Programming framework to solve a Mixed Integer Programming models.

[Charles Rosa]

To all,

I figured out the issue.  I was setting the time limit to 400 (thinking this was interpreted as seconds).

However, it was interpreted as milli-seconds....which explains why the solver stopped almost immediately.

Resetting it to 400000 resulted in an integer feasible solution.

Thanks, Charlie

On Tuesday, July 25, 2017 at 9:15:02 AM UTC-4, Charles Rosa wrote: