Computing Star Discrepancies with Numerical Black-Box Optimization Algorithms - Code and Data

This repository contains the code and data for reproducibility of the paper 'Computing Star Discrepancies with Numerical Black-Box
Optimization Algorithms'. 

The following files are included:

- TA.zip and DEM.zip: The code used for the TA and DEM algorithms respectively.

- experiment_runner: Python file which was used to run the black-box optimization algorithms on the discrepancy problems from IOHexperimenter (requires package 'ioh', version 0.3.6 or higher). This generates data in IOH-format, which is included in 'raw_data.zip'

- process_stardicr.R: R script which uses IOHanalyzer to extract the performance from the raw data into csv files for visualization. The resulting csvs are included in 'csv_with_pos' for the final results including the corresponding coordinates and 'csv_perf.zip', which contains the convergence information.

- Found_Values: The discrepancy values found by TA and DEM, separated by sampler.

- A csv file of the relative performance of each of the optimizers compared to the values found by TA is included in 'final_precision_table.csv'

- Plot_StarDiscr: the python notebook used to generate all figures, except figure 3 which was created using the IOHanalyzer GUI (iohanalyzer.liacs.nl). The full dataset is available on the website under the source 'star_discrepancy'

- Figures: some additional figures which were not included in the paper because of space constraints + higher quality versions of some of the landscape plots.