2.0.4

Minor release:

• Fix PyPi package
• Don't bundle estimator dependencies

2.0.3

• Add FeatBoost support
• Add Stability Selection support (forward-compatibility via fork)

2.0.2

Minor release.

• Adds README information to PyPi page

2.0.1

• Bugfixes
• Fix PyPi distribution

2.0.0 🚀

Completely rewritten framework, integrating wandb, hydra and sklearn for easily configurable and flexible feature selection benchmarking.

• Beautifully visualize benchmarking results inside a wandb dashboard ✨✨
• Ability to enqueue jobs in a Redis database using Hydra's RQ Launcher.
• Both classification and regression datasets supported.
• Multi-label classification and multi-output classification- and regression all supported.
• Flexible dataset loading using adapters: adapters for OpenML and wandb artifacts are included.
• Ability to dynamically define a feature importances ground-truth on any dataset - can be used to evaluate feature rankings.
• Sophisticated evaluation metrics are used to evaluate the quality of a feature ranking.

🙌🏻

1.0.0: Initial release

• Benchmark feature selection algorithms
• Complete pipeline for benchmarking: reading datasets, performing feature selection, computing evaluation metrics, plotting results
• Written from the ground up to be parallel-compatible
• SLURM interoperability (e.g. University of Groningen's Peregrine cluster)
• CLI tool for submitting jobs