AutoGMM

Automatic Gaussian Mixture Modeling in Python.

Install

pip install autogmm

Quick start

from autogmm import AutoGMM
from sklearn.datasets import make_blobs

X, _ = make_blobs(
                 n_samples=1000,
                 centers=4,
                 cluster_std=1.2,
                 random_state=0
)

labels = AutoGMM(
                 min_components=1,
                 max_components=10, # unknown K
                 criterion="bic",
                 random_state=0
).fit_predict(X)

Features

• Initializations: KMeans, Ward–Euclidean, Ward–Mahalanobis

• EM with eigenvalue flooring and covariance constraints (spherical, diag, tied, full)

• Model selection via BIC/AIC (unknown K)

• Optional spectral front-end (ASE/LSE) for nonconvex structure

• Parallel evaluation, clean API, reproducible scripts

Documentation

• API & Guides: https://github.com/neurodata/autogmm/
• Examples: examples/ (benchmarks; stress tests; runtime scaling)
• Reproducibility: scripts/reproduce.sh (regenerates all figures with fixed seeds)

Legacy & Independence

AutoGMM was originally developed in the graspologic library. As of v1.0, it is a standalone package with no dependency on graspologic.

Contributing

Issues and PRs are welcome. See CONTRIBUTING.md.

Citation

@software{autogmm,
  title   = {AutoGMM: Automatic Gaussian Mixture Modeling in Python},
  author  = {Liu, Tingshan and Athey, Thomas L. and Pedigo, Benjamin D. and Vogelstein, Joshua T.},
  year    = {2025},
  url     = {https://github.com/neurodata/autogmm}
}

License

BSD 3-Clause License. See LICENSE.