This repository contains all scripts and analysis related to experiments made for the paper "Classification and evaluation of the algorithms for vector bin packing", published in Computers and Operations Research journal and openly available on HAL.

If you need to refer to material taken from this repository, please cite our paper: Clément Mommessin, Thomas Erlebach, and Natalia Shakhlevich. "Classification and evaluation of the algorithms for vector bin packing". In Computers and Operations Research, 173, 2025.

Note that this repository also contains materials for experiments in 1 dimension, which is a work in progress.

Repository content

All instance files have been moved to the companion repository Benchmark_instances

All instance files, for the 3 benchmarks 'Panigrahy', 'New' and 'Triplet' are found in data, along with their generation script in Python.

The results_data folder contains:

• All result files, storing the solution found for every algorithm applied to every instance (files named Panigrahy_multidim_full.csv, New_multidim_full.csv and Triplet_multidim.csv)
• Summary files (named <benchmark_name>_summary_known.csv) containing, for each instance, the value of the lower bound (LB), the optimal value (OPT) if known or -1 otherwise , and the best known solution for each main family of algorithm (best_overall for all algorithms, best_IC for all item-centric algorithms, best_BC for all bin-centric algorithms, best_MB-Pairing for all MB-Pairing algorithms, and best_MB-WFD for all MB-WFD and MB-BFD algorithms).
• In folder notebooks the Jupyter notebooks used for analysis of results
• In folder plots the plots generated by the notebooks

The src folder contains the source files used to run the algorithms and generate the result files. The program relies on the Vectorpack_cpp library that implements all algorithms.