License review of the criteria
Source:vignettes/articles/criteria-license-review.Rmd
criteria-license-review.RmdTSDFGS is released under GPL (>= 3). For each criterion we checked who published it, where the original code lives, under what license, and whether a patent notice applies. This is our reading of the public record, not legal advice. Reviewed 2026-11-04.
Rules we applied
- Formulas and algorithms are ideas and methods; copyright does not protect them. We implement every criterion from the published definition in our own C++ (RcppEigen).
- We read code only when its license is GPL-3 or GPL (>= 3), as for STPGA (GPL-3) and TSODBG (GPL >= 3). We used it to check our numbers and did not copy code; no line of TSDFGS comes from these packages. Credit goes to their authors on each page.
- We did not read or copy code without a license or under restrictive terms.
- We cite the article for each criterion and link the authors’ repository or website.
Sources
| Source | License | What we did |
|---|---|---|
| STPGA (Akdemir et al.) | GPL-3 | Read; compared numerically (PEV, CD, A-, D-, E-optimality, PEVmax, G-optimality, CDmean) |
| TSODBG (Chen et al., 2024) | GPL (>= 3) | Read; compared (MSPE, CDmean(v2), GVoverall, GVaverage) |
| TrainSel (Akdemir et al.) | All rights reserved; says concepts may be covered by patents | Not read, not used |
| Code of Fernández-González et al. (2023) | No LICENSE file | Not read, not used; definitions from the open-access article |
| Core Hunter (De Beukelaer et al.) | MIT | Not used; core-collection selection is a different task |
Left out
- TrainSel algorithms: license and patent notice above.
- Core-collection measures (Core Hunter: entry-to-nearest-entry, allelic richness): they select a core collection, not a training set for prediction.
- Criteria without a published closed-form definition we could verify, for example uniform coverage and FURS clustering. They may be added later.