frsutils documentation¶
frsutils provides reusable fuzzy-rough set components and task-oriented APIs
for similarity construction, lower and upper approximations, boundary regions,
and positive-region scores.
Use the navigation to find public API contracts, execution and backend behavior, benchmark guidance, and the scientific definitions implemented by the library.
User documentation¶
- Public API: canonical imports, task-level APIs, result objects, scorer usage, and downstream-package boundaries.
- Public configuration contract: selectors, naming prefixes, registered aliases, component parameters, and model-specific usage.
- Backends: dense, blockwise, NumPy/CuPy behavior, backend metadata, and model-specific GPU claim boundaries.
- Performance Benchmarking: benchmark commands, recorded fields, paired NumPy/CuPy comparison, and safe interpretation rules.
- Reproducible reference study: real-dataset analysis, dense/blockwise equivalence evidence, environment capture, and committed results.
- Glossary: short canonical definitions for execution and fuzzy-rough terminology.
Scientific concepts¶
The concept notes are intentionally kept as separate files so each fuzzy-rough component can be reviewed independently:
Developer and release documentation¶
- Maintainer release process: version-independent release commands, validation steps, package builds, and publication checks.
- Software archiving and DOI metadata: immutable software archives, version-specific DOIs, and citation maintenance.
Documentation policy¶
Keep new documentation in one of these places unless there is a strong reason to split it:
- user-facing API, usage notes, or glossary terms:
docs/user/ - scientific concept notes:
docs/concepts/ - release, testing, and maintainer process:
docs/developer/
Avoid adding short-lived status reports or duplicate checklist files to the active docs tree. If temporary notes are needed, keep them outside active docs or remove them before release.
Community¶
Project governance and collaboration resources are maintained in the repository:
Local documentation build¶
python -m pip install -e ".[docs]"
mkdocs build --strict
mkdocs serve