README.rst

Parametric Cut Generation

Parametric Cut Generation contains the source code for the python package parametricCutGen based on the dissertation of Acadia Larsen.

parametricCutGen implements a single row optimal cut selection for Mixed Integer Programs over the domain (and restricted domains) of continuous minimal functions with at most k breakpoints.

This repository is in an alpha state. Version 0.1 coming soon!

Current Intents

• Illustrate concept of parametric cut generation and optimal cut generation as a proof of concept for MIP solvers.
• Reproducibility of experimental data using a HPC.
• Demonstrate use of passsagemath and cutgeneratingfunctionology in application; in particular illustrate application of cutting edge mathematics to application of MIPs.
• Documentation is intended support to my dissertation.

Installation

This repository is currerntly only available from source.

An installation of cutgeneratingfunctionlogy, passagemath, pplitepy, pyscipopt, scipy, and cvxpy are required. To install:

git clone https://github.com/ComboProblem/parametricCutGeneration.git
cd parametricCutGeneration
python3 -m venv /cgp-env/venv
source /cgp-env/venv/bin/activate
git clone --branch MinFunStable https://github.com/ComboProblem/cutgeneratingfunctionology.git
cd cutgeneratingfunctionology
pip install '.[passagemath]'
pip install cvxpy
pip install scipy
pip install pyscipopt
pip install pplitepy
pip install .

If you wish to run a container, an apptainer .def file is provided. The recommened build is given.:

git clone https://github.com/ComboProblem/parametricCutGeneration.git
cd parametricCutGeneration
apptainer build src/Experiments/source/Apptainer.def src/Experiments/container/cgp.sif
apptainer run src/Experiments/container/cgp.sif bash

See src/Experiments/readme.rst for details about use with a cluster.

Examples

Cut generation problems can be used in pyscipopt via optimal cut generation. In python, OptimalCut can be added in the following way.:

from parametricCutGen.optimal_cut_generation import OptimalCut
from pyscipopt import Model
model = Model()
sepa = OptimalCut(cgp_kwds={'algorithm':'bkpt_as_param', 'backend':'pplite', 'cut_score':'parallelism',  'epsilon': 1/4, 'M':1e6})
model.includeSepa(sepa, 'optima_cut', 'Optimally generated cuts using breakpoints as parameters algorithm', priority=1000, freq=1)

See src/Experiments/README.rst for a quick primier on optimal cut parameters.

License

The code is released under the GNU General Public License, version 2, or any later version as published by the Free Software Foundation.