Preregistration SF-1: mean-started descent in differential evolution on an unseen, controlled function family (predictions on valley structure, oscillation, rotation and conditioning)
Preregistration of experiment SF-1 for the manuscript "Unmasking the population mean: mean-started local descent for differential evolution on oscillatory big-valley landscapes" (in preparation). The plan, the function generator, the analysis scripts and the pre-run test log were frozen by MD5 hashes before any run with the experiment's run seeds. A new function generator varies four properties independently: valley structure (single funnel; double funnel with the Lunacek parameters), oscillation (none; single-scale Rastrigin type; multi-scale Weierstrass type with the BBOB f16 parameters, on the same lattice of minima), rotation (random orthogonal matrix or none) and conditioning (1 or 100), at D = 30 and 50: 48 cells. jSO with and without the mean-started descent (KW), 25 seeds (88501-88525), MaxFES = 10^4 D. A cell counts as a gain if the median log10 error of KW at half budget is at least 0.3 below that of jSO. Preregistered predictions: P1, gain in at least 6 of the 8 rotated single-scale cells; P2, no gain in at least 13 of the 16 multi-scale cells; P3 (control), no gain in at least 13 of the 16 cells without oscillation; no prediction for the 8 unrotated single-scale cells. Verdict: CORE HELD if P1 and P2 hold, PARTIAL if one holds, NOT HELD otherwise. Files: sf1_onkayit.tgz (MD5 b6a72358351dd99e2d6740db8b0a0043), plan kam-sf1-plani.md (MD5 8fa7a4d38e43cb395c85556ae15a9aba, in Turkish), SF1_DONDURMA_MD5.txt (hashes of all frozen files, including binaries and third-party files that are not redistributed). Related preregistration: PB-1, doi:10.5281/zenodo.23940212.
Authors
- Erdinç Şahin (ORCID: https://orcid.org/0000-0002-9740-599X)
Institutions
- Giresun University (TR)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-05
- DOI
- https://doi.org/10.5281/zenodo.23980495
- Primary Topic
- Metaheuristic Optimization Algorithms Research
- Type
- article
- Field-Weighted Citation Impact
- 0.00