Search code examples
mathematical-optimizationscipsat

Is it possible to retain information after solving a problem and reusing the same information to solve the same problem but with different settings?


My problems are large SAT problems. In SCIP7.0.0, there are many branching rules to choose from. If I send in a problem to SCIP and SCIP solves the problem, is it possible to use the learnt information throughout the solving process, say cutting planes, conflict constraints and variable branching history to solve the same problem with a different method?

For example, I send in a SAT instance of 7500 variables with 450,000 constraints and optimize using default settings. Would it be possible to use the information stored in SCIP (as above), on the same problem using fullstrong branching?


Solution

  • I know this answer comes late, but why do you want to do this? If you want to compare the different branching rules, then using information you have learned in a previous solve will affect the solving behavior of the resolve with a different branching rule significantly.