Loading...
Loading...
Contributors:
Dataset URI:
10.5281/zenodo.23502467Version 2.0.0 contains repaired historical CEC05, CEC14 and SOCO aggregates and directly sampled native MIXTURE observations. Historical tables retain 132 registry records, 128 usable records and 12,800 feature observations. Aggregation uses featurewise finite masks and population standard deviation (ddof=0); historical implicit extraction bounds and unrecorded R seeds remain unchanged. Historical BFF groups are excluded. Native sampling covers all 50 MIXTURE functions at dimensions 10 and 50 with RRMset 7, using scrambled Sobol designs of 100 times dimension points, explicit registry bounds, separate seed roles, R 4.3.3 and flacco 1.8 with pinned dependencies. The preferred 100-observation view uses an upper-extended IC epsilon grid under explicit-bounds-seeded-one-cell-ic-extended-v3, preserving original tours, seeds and saved objective values. It has 6,400/6,400 finite predictor cells. The original default-grid view is retained with two missing settling sensitivities. Eight undefined one-cell LIMO columns remain excluded from the 64-predictor allowlist. An independent-design pilot adds 64 observations across 16 selected native identities. Each selected identity has five designs including its preserved baseline; the remaining 84 have one. There are 164 distinct native observations overall, with 16 duplicated baseline rows in the pilot tables. Descriptive feature spread is supplied under both IC protocols; the preferred 80-row pilot has 5,120/5,120 finite predictor cells. Coordinate, objective and seeded feature replay checks pass. The original collection used 600,000 objective evaluations including replay, and the repeat pilot used 384,000, total 984,000. IC diagnosis and re-extraction used zero objective evaluations. No imputation, global scaling or automatic performance join is performed. Historical generating executables remain unattested. Native p6 and p42 retain fixed internal CEC05 noise streams; the pilot does not establish independent noise repetitions or separate sampling and extraction RNG variance. Most source-declared optima lack independent certificates. Public companion archives provide saved X,y, extraction evidence and code for offline reproduction. Third-party evaluator binaries and benchmark support payloads are excluded, with hashes and source references retained. Data use CC-BY-4.0, project/framework code retains MIT notices and flacco-derived portions retain BSD 2-Clause notices. Implementation source: https://github.com/clcamachov/metafor-rl/commit/32bf087039c5f131393398e8cb798374c43aade6
Keyword(s):
algorithm configuration; algorithm selection; benchmark functions; black-box optimization; continuous optimization.
Grants and Funding:
No Grants or Funding Data Available
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)