Selin Bayramoglu, Georgia Institute of Technology
Nick Sahinidis, Georgia Institute of Technology
A repository to learn simple and interpretable models for cutting plane selection within the open-source solver SCIP.
Please see COPYRIGHT.md .
This project uses the open-source solver SCIP 8.0.0. To install it, run
conda install -c conda-forge scip=8.0.0
Our base instance is or-lib/10_0_1_w, a unit commitment problem from the UnitCommitment.jl package ( [XaKaYuHeQi04], [ORLIB], [FrGe06]). It involves an hourly time horizon of 24 hours, a single bus and ten generators. For our experiments, we generate a total of 50 instances by multiplying each load of the base instance with (1 + x), where x ~ N(0, 0.05^2) and independently distributed.
All instances reside in the instances/ folder. The split of problems into training, validation, test and evaluation sets are provided in the problem-lists/ folder.
To collect data for the training dataset, run
bash part1-collect-data.sh train .
For the validation and test datasets, replace the argument train with valid and test, respectively.
To build datasets, run
bash part2-create-datasets.sh 0 .
Changing the argument will produce a different training dataset by selecting a different subset of all collected training observations.
To build a models with one of the targets (e.g., logScore) in targets.json, run
bash part3-build-model.sh logScore .
To evaluate the learned models, SCIP's default cut selector or the bound improvement target, select a cut selector (e.g., logScore) and run
bash part4-evaluate-cut-selector.sh logScore .
If no solution file is provided, set NOSOL to 1 in the script.
To summarize and compare results, run
bash summarize_results.sh .
This script produces statistics such as mean solving time and integrality gap closed (IGC) over the evaluated instances.
For runs with NOSOL=1, the solver is not likely to find a primal solution at the root, therefore the IGC will be 0. These instances can be compared by the final dual bounds, where higher bounds indicate more progress.
[XaKaYuHeQi04] Alinson S. Xavier, Aleksandr M. Kazachkov, Ogün Yurdakul, Jun He, Feng Qiu, "UnitCommitment.jl: A Julia/JuMP Optimization Package for Security-Constrained Unit Commitment (Version 0.4)". Zenodo (2024). DOI: 10.5281/zenodo.4269874.
[ORLIB] J.E.Beasley. "OR-Library: distributing test problems by electronic mail", Journal of the Operational Research Society 41(11) (1990). DOI: 10.2307/2582903
[FrGe06] A. Frangioni, C. Gentile. "Solving nonlinear single-unit commitment problems with ramping constraints" Operations Research 54(4), p. 767 - 775, 2006. DOI: 10.1287/opre.1060.0309