Simple Self-Organizing Maps in Python


Keywords
kohonen, self-organizing, maps, clustering, dimension-reduction, som, dimensionality-reduction, python, self-organizing-map
License
MIT
Install
pip install simpsom==3.0.0b0

Documentation

SimpSOM (Simple Self-Organizing Maps)

DOI PyPI version Documentation Status example workflow codecov

Version 3.0.0

Simple Self-Organizing Maps (SimpSOM) is a lightweight Python 3 library to train SOM. It offers an efficient way of training SOM in Python while keeping its implementation simple and easy to read.

Version 3 is a rewriting focusing on performance.

Installation

simpsom can be downloaded from PyPI with

pip install simpsom

To install the latest (unreleased) version you can download it from this repository by running

git clone https://github.com/fcomitani/simpsom
cd simpsom
python setup.py install

Dependencies

Core dependencies:

  • numpy
  • scikit-learn
  • matplotlib

If available, CuPy can be used to run simpsom on the GPU. CuML is also optional, but will allow you to run clustering on the GPU as well.

For a full list see requirements.txt

Example of Usage

Running simpsom is easy. After setting up a network by providing size and tiling style, train it with the train method.

import simpsom as sps

net = sps.SOMNet(20, 20, data, topology='hexagonal', 
                init='PCA', metric='cosine',
                neighborhood_fun='gaussian', PBC=True,
                random_seed=32, GPU=False, CUML=False,
                output_path="./")

net.train(train_algo='batch', start_learning_rate=0.01, epochs=-1, 
    batch_size=-1)

The trained map can be saved to disk.

net.save_map("./trained_som.npy")

The results can be inspected with a variety of plotting functions.

net.plot_map_by_difference(show=True, print_out=True)
net.plot_projected_points(projected_data, color_val=[n.difference for n in net.nodes_list],
        project=False, jitter=False, 
        show=True, print_out=False)

Detailed documentation, API references and tutorials can be found here.

Who is using SimpSOM

Here are some of the research works that use SimpSOM:

Postema, J. T. (2019). Explaining system behaviour in radar systems (Master's thesis, University of Twente).

Lorenzi, C., Barriere, S., Villemin, J. P., Dejardin Bretones, L., Mancheron, A., & Ritchie, W. (2020). iMOKA: k-mer based software to analyze large collections of sequencing data. Genome biology, 21(1), 1-19.

Saunders, J. K., McIlvin, M. R., Dupont, C. L., Kaul, D., Moran, D. M., Horner, T., ... & Saito, M. A. (2022). Microbial functional diversity across biogeochemical provinces in the central Pacific Ocean. Proceedings of the National Academy of Sciences, 119(37), e2200014119.

Contributions

Contributions are always welcome. If you would like to help us improve this library please fork the main branch and make sure pytest pass after your changes.

Citation

When using this library for your work, please cite the appropriate version from Zenodo

Federico Comitani. (2022). SimpSOM (v2.0.2). Zenodo. https://zenodo.org/record/7187332