Originally posted by: ilovetoyota

Every time I call MIPOPT from C, I can see the following output from CPLEX

MIP Presolve eliminated 5775 rows and 28000 columns.
Reduced MIP has 4500 rows, 8750 columns, and 33775 nonzeros.
Presolve time = 0.45 sec.

It seems that MIP Presolve did an excellent job in reducing the problem size, and eliminate those redundant rows and columns. May I know what are the heuristics used in MIP Presolve? Is there any paper or article introduce this?

It would be super powerful to combine those CPLEX heuristics with external heuristics if we know what CPLEX do exactly.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

There are lots of different algorithms involved in CPLEX presolve. Actually, it is one of the most complex parts of the whole algorithm.
You can read about basic MIP presolving techniques in Savelsbergh, "Preprocessing and Probing Techniques for Mixed Integer Programming Problems", ORSA Journal on Computing (6), No. 4, 1994.
More details can be found in Achterberg, "Constraint Integer Programming", PhD Thesis, Technical University Berlin, 2007, http://opus.kobv.de/zib/volltexte/2009/1153/ Chapter 10.

Hope this helps.

Tobias
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: ilovetoyota

thanks
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: ilovetoyota

Tobias, is your thesis written in English?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Yes, it is. Just download it from the URL that I posted.
#CPLEXOptimizers
#DecisionOptimization