DMS-SI-Mix: Reproducibility Package

Reproducibility package for the two-space scale-invariant Direct Multisearch method DMS-SI-Mix, a derivative-free method for mixed-variable multiobjective optimization. This release contains the data, configuration files, run archives, per-poll logs, metric tables and validation scripts supporting the experiments reported in the accompanying article: - data/MORAP_NM/ : the no-mixing multiobjective reliability allocation instance, the exhaustive enumeration that produces its exact Pareto front, and the frozen front itself (578 nondominated points), with the data-freeze record; - configurations/ : the parameter files and categorical permutation schedules defining every reported run; - results/archives/ : the final MATLAB archives of each run, each carrying the nondominated list, the step-size and coverage state, and a per-poll log (State.PollLog) recording trials generated and kept, cache hits, new black-box evaluations, acceptances and categorical coverage; - results/metrics/ : the scripts and output tables for the hypervolume, IGD, recall, precision and anytime analyses, including the categorical- neighborhood ablation and its initial-design sensitivity study; - nsga2_reference/ : NEW IN 1.1.0. The external mixed-variable NSGA-II reference point on MORAP-NM reported in Section 5.4 and Appendix A.4 of the article: nine configurations x 30 seeds x 6 checkpoints, 1620 result rows, measured against the same frozen 578-point exact front under the same budget semantics as the DMS-SI-Mix poll variants. Python 3.11.15 with pymoo 0.6.2, self-contained and independent of the MATLAB solver. Includes the resumable harness, the aggregation and oracle-envelope scripts, the paired initialization control, the anytime-curve figure, and an integrity audit that exits non-zero on failure. Note that this component uses a Hypervolume reference computed from the exact front alone, which is independent of any run and differs from the union-based reference of Table 6 of the article; recall, precision and IGD are reference-independent and are directly comparable as published; - MANIFEST.txt : SHA-256 hashes for every file in the release, together with the MATLAB version used to produce them (R2025b, 25.2.0.2998904, PCWIN64). This version does not contain the MATLAB implementation of the solver. The solver is deposited as version 2.0.0 of this record upon acceptance of the accompanying article. Until then it is available to reviewers on request through the journal editor.