Changelog

1.8.3 (2025)

• Declared missing runtime dependencies (numpy, scipy, matplotlib, networkx) in pyproject.toml so pip install NNSOM installs them automatically.

• Overhauled Sphinx documentation: RTD theme, full API reference generated from docstrings, Quick Start and Installation guides, Changelog page.

• Added NumPy-style docstrings to all previously undocumented utility functions in utils.py.

1.8.2 (2024)

• Removed author email from package metadata.

• Minor documentation and packaging improvements.

1.8.1 (2024)

• Patch release: corrected wheel and source distribution builds.

• Added Forest example dataset under examples/Tabular/Forest/.

1.8.0 (2024)

• Restructured package layout to src/NNSOM standard.

• Added SOMPlots visualization class with interactive plot support via plot().

• Added GPU-accelerated SOMGpu with automatic CuPy detection; SOMPlots falls back to NumPy when CuPy is unavailable.

• New quality metrics: quantization_error(), topological_error(), distortion_error().

• Utility functions for cluster statistics: get_cluster_avg(), majority_class_cluster(), get_perc_cluster(), and more.

• Sub-clustering support via SOM.sub_som.

• Pickle-based model persistence: save_pickle() / load_pickle().

• Iris dataset example notebooks included.