About
The project
Section titled “The project”AMM (Adaptive Multi Mode) is a derivative-free optimizer for continuous black-box problems. It reads the function with cheap probes and routes the evaluation budget to a classical algorithm. It grew out of an earlier optimizer, ATM (Adaptive Two Mode), whose sampler is one stage inside AMM. The code is Python, depends on NumPy and SciPy, and is developed with a preregistered protocol that the research log describes.
This site states every result next to its status: final, in-sample development data, or pending. The limits page lists where AMM loses.
Licence
Section titled “Licence”Licensed under Apache-2.0. The package version in the repository is 0.3.0, and the PyPI release is 0.3.0 on PyPI (0.2.0 yanked).
Contact
Section titled “Contact”- General: pendingpending
- Security reports: pendingpending. The repository has a
SECURITY.md. - Source repository: https://github.com/BjBodner/amm-optimizer (public). The research archive (preregistrations, FT1 data, id ledger, tagged history) is planned as a Zenodo deposit, DOI pendingpending.
How this site and the paper were written
Section titled “How this site and the paper were written”AI tools supported the drafting of the paper, the site text, tables and figures, and helped with code and literature search: pendingpending. The author makes the decisions and does the final edit pass. Numbers on the site are not typed into the text: each comes from one data file and carries its status (final, in-sample or pending). AI support does not replace the author’s checking: responsibility for the content stays with the author.