A posteriori certificates of Pareto optimality in finite multi-objective optimisation: companion Python verification kit

Status note — 23 June 2026. This record documents the companion software prepared for the former Journal of Global Optimization submission entitled “A posteriori certificates of Pareto optimality in finite multi-objective optimisation”. It is retained solely for provenance and reproducibility of that submission. It is not the companion artefact for the substantially reconstructed manuscript on sparse exact certificates for weighted-sum unsupportedness. The successor software release is archived under a separate Zenodo record and DOI. This record archives the companion Python verification kit for the manuscript “A posteriori certificates of Pareto optimality in finite multi-objective optimisation”. The kit reproduces the computational checks and synthetic experiments reported in the manuscript. It contains terminal-only Python code, tests, synthetic instance generators and figure-generation scripts. It does not contain Jupyter notebooks, Spyder project files, personal data, institutional data or third-party proprietary datasets. The archived code supports the following components of the manuscript: 1. verification of finite linear/Lagrangian certificates over declared verification domains;2. distinction between weak and strict certificate modes;3. augmented Chebyshev certification on finite objective images;4. stability-radius checks based on strict certificate gaps;5. candidate-relative fractional-objective checks;6. delayed constraint-generation certificate search on synthetic bi-objective knapsack and assignment instances;7. reproduction of the figure and JSON result files used by the manuscript. The record is intended to support scholarly inspection and reproducibility assessment of the associated manuscript. It is not released as open-source software. The files are governed by the Proprietary Research-Reproducibility Licence v1.0 included in the archive. The licence permits local execution solely to reproduce the reported synthetic experiments and prohibits redistribution, modification, sublicensing, commercial use, derivative works, incorporation into other software, republication and model-training or benchmark-mining uses without prior written permission from the copyright holder. To reproduce the reported outputs, download the archive, create a Python environment, install the dependencies listed in requirements.txt and run: python experiments/run_all.pypytest tests -q The synthetic instances are fully determined by the seeds reported in the manuscript and in the experiment scripts. No human-subject data, personal data or external research datasets are used.