Nimfa

Nimfa is a Python module that implements many algorithms for nonnegative matrix factorization. Nimfa is distributed under the BSD license.

The project was started in 2011 by Marinka Zitnik as a Google Summer of Code project, and since then many volunteers have contributed. See AUTHORS file for a complete list of contributors.

It is currently maintained by a team of volunteers.

[News:] Scikit-fusion, collective latent factor models, matrix factorization for data fusion and learning over heterogeneous data.

[News:] fastGNMF, fast implementation of graph-regularized non-negative matrix factorization using Facebook FAISS.

• Official source code repo: https://github.com/marinkaz/nimfa
• HTML documentation (stable release): http://ai.stanford.edu/~marinka/nimfa
• Download releases: http://github.com/marinkaz/nimfa/releases
• Issue tracker: http://github.com/marinkaz/nimfa/issues

Nimfa is a fantastic library for non-negative matrix factorization in Python. I don’t have the expertise to optimize the underlying algorithm, but I’ve identified compatibility issues with NumPy 2.0 that are causing errors.The primary goal of this fork is to adjust syntax to accommodate Python syntax updates and account for deprecated functions in NumPy 2.0. e.g.: np.mat to np.array

All modifications were auto-generated by GitHub Copilot in VS Code.

It works in my conda env with Python 3.11.11 with Numpy 2.1.0 and Scipy 1.15.2. (2025.08.12)

Install

pip install git+https://github.com/omega-sz/nimfa.git

Nimfa is tested to work under Python 2.7 and Python 3.4. The required dependencies to build the software are NumPy >= 1.7.0, The required dependencies to build the software are NumPy >= 2.0.0, SciPy >= 0.12.0.

For running the examples Matplotlib >= 1.1.1 is required.

This package uses setuptools, which is a common way of installing python modules. To install in your home directory, use:

python setup.py install --user

To install for all users on Unix/Linux:

sudo python setup.py install

For more detailed installation instructions, see the web page http://ai.stanford.edu/~marinka/nimfa.

Alternatively, you may also install this package using conda:

conda install -c conda-forge nimfa

Run alternating least squares nonnegative matrix factorization with projected gradients and Random Vcol initialization algorithm on medulloblastoma gene expression data:

>>> import nimfa
>>> V = nimfa.examples.medulloblastoma.read(normalize=True)
>>> lsnmf = nimfa.Lsnmf(V, seed='random_vcol', rank=50, max_iter=100)
>>> lsnmf_fit = lsnmf()
>>> print('Rss: %5.4f' % lsnmf_fit.fit.rss())
Rss: 0.2668
>>> print('Evar: %5.4f' % lsnmf_fit.fit.evar())
Evar: 0.9997
>>> print('K-L divergence: %5.4f' % lsnmf_fit.distance(metric='kl'))
K-L divergence: 38.8744
>>> print('Sparseness, W: %5.4f, H: %5.4f' % lsnmf_fit.fit.sparseness())
Sparseness, W: 0.7297, H: 0.8796

@article{Zitnik2012,
  title     = {Nimfa: A Python Library for Nonnegative Matrix Factorization},
  author    = {Zitnik, Marinka and Zupan, Blaz},
  journal   = {Journal of Machine Learning Research},
  volume    = {13},
  pages     = {849-853},
  year      = {2012}
}

• Data fusion by matrix factorization: http://dx.doi.org/10.1109/TPAMI.2014.2343973
• Jumping across biomedical contexts using compressive data fusion: https://academic.oup.com/bioinformatics/article/32/12/i90/2240593
• Survival regression by data fusion: http://www.tandfonline.com/doi/abs/10.1080/21628130.2015.1016702
• Gene network inference by fusing data from diverse distributions: https://academic.oup.com/bioinformatics/article/31/12/i230/216398
• Fast optimization of non-negative matrix tri-factorization: https://doi.org/10.1371/journal.pone.0217994

• A comprehensive structural, biochemical and biological profiling of the human NUDIX hydrolase family: https://www.nature.com/articles/s41467-017-01642-w
• Gene prioritization by compressive data fusion and chaining: http://dx.doi.org/10.1371/journal.pcbi.1004552
• Discovering disease-disease associations by fusing systems-level molecular data: http://www.nature.com/srep/2013/131115/srep03202/full/srep03202.html
• Matrix factorization-based data fusion for gene function prediction in baker's yeast and slime mold: http://www.worldscientific.com/doi/pdf/10.1142/9789814583220_0038
• Matrix factorization-based data fusion for drug-induced liver injury prediction: http://www.tandfonline.com/doi/abs/10.4161/sysb.29072
• Collective pairwise classification for multi-way analysis of disease and drug data: https://doi.org/10.1142/9789814749411_0008

• Hidden Genes: Understanding cancer data with matrix factorization, ACM XRDS: Crossroads: https://dl.acm.org/citation.cfm?id=2809623.2788526 [Jupyter Notebook]