piquepq / ReDUCE

ReDUCE: Reformulation of Mixed Integer Programs using Data fromUnsupervised Clusters for Learning Efficient Strategies

ReDUCE

ReDUCE: Reformulation of Mixed Integer Programs using Data fromUnsupervised Clusters for Learning Efficient Strategies

This repository contains code for ReDUCE: Reformulation of Mixed Integer Programs using Data fromUnsupervised Clusters for Learning Efficient Strategies by Xuan Lin, Gabriel I. Fernandez, Dennis W. Hong

Construction ongoing.

To collect data, run:

python complete_pipeline.py

To train learning agents, use the jupyter notebook:

jupyter notebook book_problem_train.ipynb