Regression and Classification Set + Recommended Loading Scripts

What's Changed

Release Notes: https://srlearn.github.io/relational-datasets/changelog/#v040-2022-11-03

• ✨ Bump datasets to v0.0.6, Add california_housing and roofworld20 by @hayesall in #28
• 🚧 Bump codecov-action to v3 by @hayesall in #29
• ✨ Dataset loading recommendations by @hayesall in #30
• 🔧 Set pythonpublish.yml to use pypi-publish by @hayesall in #31
• 📝 Add changelog entry for v0.4.0 by @hayesall in #32

Full Changelog: v0.3.0...v0.4.0

Unique names assumption and avoiding type conflicts

What's Changed

• 🐛 Append variable name to constant in from_numpy by @hayesall in #25
• 📝 Documentation updates / fix default handler to use Python by @hayesall in #27
• 📝 Update notebooks to align with variable name changes by @hayesall in #26

Full Changelog:

• v0.2.2...v0.3.0
• https://srlearn.github.io/relational-datasets/changelog/#v030-2022-06-20

Multiclass support

Software Changes:

• Add multiclass support to convert.from_numpy

Documentation:

• Add notebook with overview on converting multiclass vector datasets

Testing:

• Add tests for convert.from_numpy

Drug Interactions and Toy Machines

Software Changes

• Add drug_interactions and toy_machines datasets
• Add v0.0.5 as the latest srlearn/datasets release

v0.2.0

Software Changes:

• Add convert module with from_numpy implementation to convert binary
  classification and regression datasets based on ordinal encodings.
• Fix type annotations in relational_datasets.request
• Fix type annotations in relational_datasets.types

Documentation:

• Add tutorial for converting vector/propositional datasets to relational
• Add mkdocs dependency: pymdownx.tasklist
• Add binder and colab launch badges to Jupyter notebook tutorials

Testing:

• Add lgtm build step + README badge
• Add codecov build step + README badge
• Add numpy>=1.20.0 as an optional setup target, and test against it
  (this is the earliest version of numpy where type annotations for mypy
  seem to be consistently available)

v0.1.1 - Split loading and deserializing / Project Documentation

Software Changes:

• Bump default dataset version: v0.0.3 → v0.0.4.
• Between v0.0.3 and v0.0.4 of srlearn/datasets, all zipfiles now have the version number appended
  (e.g. toy_cancer_v0.0.4.zip). Add logic to request the correct filename from GitHub.
• Add deserialize_zipfile function, split out code for pulling zipfile content from the load method.
• Add private _make_file_path function to handle where zipfiles are stored on a user's filesystem.
• Move RelationalDataset type into relational_datasets/types.py
• Fix hayesall/relational_datasets → srlearn/relational_datasets in setup.py
• Clarify typing.Optional in function signatures where default file paths are allowed.
• Add __version__ to the main __init__, so print(relational_datasets.__version__) is valid.

Documentation:

• Add mkdocs builds with each push to the main branch.
• Add requirements_dev.txt with requirements to build documentation.
• Add docs/build._docs.py to build a Downloads page and an overview of each dataset pulled from the srlearn/datasets repository.
• Add docs/notebooks/ directory for literate tutorials
  • Add 00_loading_and_fetching.ipynb
• Pages for functions and types:
  • types.RelationalDataset
  • request.deserialize_zipfile
  • request.fetch
  • request.latest_version
  • request.load

🔖 Initial Release

Release basic specification:

• load(name: str, version: str = "", fold: int = 1) -> Tuple[RelationalDataset, RelationalDataset]: Load examples and facts for a dataset/version/fold
• fetch(name: str, version: str = "") -> str: Download and cache an archive locally
• get_data_home(data_home=None) -> str: Get the path to the cache directory
• clear_data_home(data_home=None) -> None: Remove the cache directory
• latest_version() -> str: Check latest version of datasets on GitHub