Benchmarking Predict-then-Optimize (PtO) Problems

About

This repository provides a comprehensive framework for benchmarking Predict-then-Optimize (PtO) problems using Decision-Focused Learning (DFL) approaches. PtO problems involve making predictions that are used as input to downstream optimization tasks, where traditional two-stage methods often lead to suboptimal solutions. DFL addresses this by training machine learning models that directly optimize for the downstream decision-making objectives.

This repository contains the implementation for the paper (Accepted to Journal of Artificial Intelligence Research (JAIR)):

Mandi, J., Kotary, J., Berden, S., Mulamba, M., Bucarey, V., Guns, T., & Fioretto, F. (2024). Decision-focused learning: Foundations, state of the art, benchmark and future opportunities. Journal of Artificial Intelligence Research, 80, 1623-1701. DOI: 10.1613/jair.1.15320

If you use this code in your research, please cite:

@article{mandi2024decision,
  title={Decision-focused learning: Foundations, state of the art, benchmark and future opportunities},
  author={Mandi, Jayanta and Kotary, James and Berden, Senne and Mulamba, Maxime and Bucarey, Victor and Guns, Tias and Fioretto, Ferdinando},
  journal={Journal of Artificial Intelligence Research},
  volume={80},
  pages={1623--1701},
  year={2024},
  doi={10.1613/jair.1.15320}
}

Installation

Prerequisites

• Python 3.7.3 (recommended)
• pip or conda package manager

Option 1: Using venv (Recommended)

1. Create and activate a virtual environment:

python3 -m venv benchmarking_env
source benchmarking_env/bin/activate

2. Upgrade pip:

pip install --upgrade pip

3. Install required packages:

pip install -r requirements.txt

Option 2: Using Conda

1. Install Conda by following the official installation guide

2. Create and activate the environment:

# Create environment
conda env create -n benchmarking_env --file environment.yml

# Activate on Linux/macOS
conda activate benchmarking_env

# Activate on Windows
source activate benchmarking_env

Running Experiments

Navigate to the corresponding experiment directory to run specific benchmarks.

Contributing

Feel free to open issues or submit pull requests if you find any problems or have suggestions for improvements.