Skip to content

Contributing

This project follows the Contributor Covenant. By taking part, you agree to that code of conduct.

This is a curated list, not an archive. A keyword match on "Bayesian optimization" is not enough.

The GitHub README is what awesome-lint checks. Edit README.md, then run python scripts/sync_docs.py so the docs pages match.

What belongs here

Papers must introduce a BO method, survey, tutorial, or benchmark that the community still uses. Typical yes: a new acquisition function, a high-dimensional algorithm, a constrained-BO formulation, a library paper.

If you have a domain paper that uses BO (materials, communications, robotics, drugs, and so on) but does not introduce a BO method, it belongs under Applications, not Papers. Applications is pointer papers and domain lists. A single case study that just runs BoTorch, Ax, or Optuna belongs in a domain list such as the materials list.

Software must be maintained (recent commits or releases), documented, and actually used for BO. Unmaintained tools go under Unmaintained. Related HPO tools that are not GP-BO (TPE, Hyperband-only, and similar) go under Related HPO tools, and the description must say so.

Preprints must have Bayesian optimization / BayesOpt in the title, and BO must be the contribution. The weekly bot only proposes title matches. Humans still merge.

How to add something

  1. Open an issue with the matching template (paper, software, or resource), or open a pull request.
  2. Put the entry in the right section and in the right place (see Where to put it). Do not append at the bottom of a subsection unless nothing nearby fits.
  3. Use this format, and nothing else:
- [Name](https://example.com) - Short why-it-belongs sentence.

The description starts with an uppercase letter and ends with a period. 4. Prefer a stable landing page (proceedings, OpenReview, arXiv abs, project site) over a raw PDF when both exist. 5. Do not duplicate a URL that is already in the README. 6. One pull request per addition, unless you are doing a small batch of clearly related items. 7. Run python scripts/sync_docs.py so the docs site matches the README.

Where to put it

Papers. Within each subsection, newest year first. Same year: alphabetical by title. Two exceptions: Foundations is a reading path, so oldest first. Surveys and Tutorials keeps the two canonical intros at the top, then newest first for the rest.

Recent Preprints is newest first. After acceptance, move the entry into the matching Papers subsection by year. Do not leave it at the end of that subsection.

Software is not ordered by year. Place a library next to the closest existing tools: same job, same stack, or the one people already compare it to. Example: BayBE sits next to BoFire because both are industrial DOE/BO libraries. Do not dump new entries at the bottom of the list.

Weekly arXiv pull requests

scripts/update_papers.py may open a PR that only adds Recent Preprints in README.md, then runs scripts/sync_docs.py. Review it:

  • Drop application papers even if the title contains "Bayesian optimization".
  • Keep method papers. After acceptance, insert them into the matching Papers subsection by year and remove them from Recent Preprints.
  • Never merge a PR that writes into the curated Papers sections automatically.

Docs site

python scripts/sync_docs.py
pip install -r requirements-docs.txt
mkdocs serve

Do not edit generated list pages (docs/books.md, docs/papers.md, and similar) by hand. The site deploys to GitHub Pages on every push that touches docs/ or mkdocs.yml.

Reporting a problem

Open an issue if a link is dead, a year/venue is wrong, a package is unmaintained, or an entry is off-topic. PRs that just fix that are welcome.