{"name": "Parametric Process", "package_name": "parametric_process", "version": "3.2.0", "experimental": false, "qgis_min": "3.28.0", "qgis_max": "4.99.0", "downloads": 283, "uploaded_by": "geo140195philo", "upload_datetime": "2026-09-29T15:04:54.452637", "changelog": "## [3.2.0] - 2026-09-29\r\n\r\n### Fixed\r\n\r\n- **NSGA-III was NSGA-II under another name.** Das-Dennis reference points\r\n  were generated and never used; the last front was filled by crowding\r\n  distance. Survivor selection now follows Deb & Jain (2014): unit-free\r\n  normalisation (ideal point, ASF extreme points, hyperplane intercepts with\r\n  a nadir fallback), association with the nearest reference line, and niche\r\n  counts that include survivors from earlier fronts. Parents are mated at\r\n  random, as in the paper. Results of earlier \"NSGA-III\" runs were NSGA-II\r\n  with genotype repair.\r\n- **MOEA/D compared raw objectives.** The Tchebycheff function now divides\r\n  by each objective's current range, so an objective in kg CO2e no longer\r\n  outweighs one in percent by six orders of magnitude; zero weights are\r\n  floored at 1e-6.\r\n- **Sol-air temperature was a constant** (66.1 degC for every design). It\r\n  now follows the ASHRAE sol-air relation with the roof's albedo.\r\n- README and showcase no longer describe the wind index or particles as CFD.\r\n\r\n### Added\r\n\r\n- Algorithm-property tests: NSGA-III niching against earlier-front\r\n  survivors (fails for crowding-distance selection), unit-free\r\n  normalisation, unit-free MOEA/D ranking, and metric response to design.", "external_deps": null, "download_url": "https://plugins.qgis.org/plugins/parametric_process/version/3.2.0/download/"}