Skip to content

Numerical tests for Section 5.3 in "Feasible rounding based strategies in branch-and-bound methods for mixed-integer optimization".

Notifications You must be signed in to change notification settings

schwarze-st/FRA_BB

Repository files navigation

FRA_BB: Numerical tests for Section 5.3 in "Feasible rounding based strategies in branch-and-bound methods for mixed-integer optimization"

We integrate feasible rounding approaches and diving ideas into the SCIP framework via pyscipopt. The following versions were used for our experiments:

The method

The diving procedure is implemented in fra_heur.py as instance of the class Heur in PySCIPOpt.

Reproducing our results

There are scripts for all experiments and data we generated for the publication. The jupyter notebook in results/notebook/results_nb.ipynb extracts all relevant data from the generated output-files

1. Filtering the test bed: filter_instances.py

See results/FilterInstances.ipynb

2. Running SCIP with fra_heur.py on the test bed: diving_analysis.py

We run SCIP with fra_heur.py on the obtained test bed with 128 instances with up to 5 diving rounds and 30 minutes maximum run time. Results are displayed in results/Evaluation.ipynb.

3. Running SCIP with plain settings: SCIP_plain.py

We run SCIP without fra_heur.py for 32 instances where it found best solutions until the solution is better than the one found with fra_heur.py. Results are displayed in results/Evaluation.ipynb.

About

Numerical tests for Section 5.3 in "Feasible rounding based strategies in branch-and-bound methods for mixed-integer optimization".

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published