PredFL (Learning Augmented Online Facility Location)

This repository contains the source code, datasets and solutions related to a paper submited to NeurIPS 2021.

1. dataset contains the datasets (CoverType, US Census) used to run the experiments for the algorithm presented in the paper. The synthetic dataset was created by randomly sampling 2k points on a grid.
2. solutions contains the (optimal) offline linear solutions for each dataset. In order to calculate the optimal solution, we solved the LP-relaxation of facility location using Gurobi version 9, and using deterministic rounding we obtained a 6approximate integer solution.
3. notebooks contains three jupyter notebooks corresponding to each dataset.

If you have any question on the algorithm implemented, please refer to said article.