ctlab / ITMO_FS

What it does

• Rich set of algorithms. supervised / unsupervised filters, wrappers, embedded, hybrid and ensemble methods in one library.
• Scikit-learn compatible API. fit, transform, fit_transform and easy integration into sklearn pipelines.
• Composable filters. separate “measure” and “cutting rule” components let you implement custom strategies (e.g. thresholds or top-k).
• Dense and sparse data support. works with NumPy arrays, pandas DataFrames and SciPy sparse matrices.
• Research background. algorithms are based on well-known methods from the literature and used in research projects.