pyunicorn (Unified Complex Network and RecurreNce
analysis toolbox) is an object-oriented Python package for the advanced analysis
and modeling of complex networks. Beyond the standard measures of complex
network theory (such as degree, betweenness and clustering coefficients), it
provides some uncommon but interesting statistics like Newman's random walk
betweenness. pyunicorn also provides novel node-weighted (node splitting invariant)
network statistics, measures for analyzing networks of interacting/interdependent
networks, and special tools to model spatially embedded complex networks.
Moreover, pyunicorn allows one to easily construct networks from uni- and
multivariate time series and event data (functional/climate networks and
recurrence networks). This involves linear and nonlinear measures of
time series analysis for constructing functional networks from multivariate data
(e.g., Pearson correlation, mutual information, event synchronization and event
coincidence analysis). pyunicorn also features modern techniques of
nonlinear analysis of time series (or pairs thereof), such as recurrence
quantification analysis (RQA), recurrence network analysis and visibility
graphs.
pyunicorn is fast, because all costly computations are performed in
compiled C code. It can handle large networks through the
use of sparse data structures. The package can be used interactively, from any
Python script, and even for parallel computations on large cluster architectures.
For information about individual releases,
see our CHANGELOG and CONTRIBUTIONS.
pyunicorn is BSD-licensed (3 clause).
Please acknowledge and cite the use of this software and its authors when results are used in publications or published elsewhere. You can use the following reference:
J.F. Donges, J. Heitzig, B. Beronov, M. Wiedermann, J. Runge, Q.-Y. Feng, L. Tupikina, V. Stolbova, R.V. Donner, N. Marwan, H.A. Dijkstra, and J. Kurths. "Unified functional network and nonlinear time series analysis for complex systems science: The pyunicorn package". Chaos 25, 113101 (2015), doi:10.1063/1.4934554, Preprint: arxiv.org:1507.01571 [physics.data-an].
The development of pyunicorn has been supported by various funding sources,
notably the German Federal Ministry for Education and Research (projects GOTHAM and CoSy-CC2), the Leibniz Association (projects ECONS and DominoES),
the German National Academic Foundation,
and the Stordalen Foundation via the
Planetary Boundary Research Network (PB.net) among
others.
Stable releases can be installed directly from the Python Package Index (PyPI):
$> pip install pyunicorn
Alternatively, source distributions can be downloaded from the GitHub Releases.
On Windows, please first install the latest version of the Microsoft C++ Build Tools, which is required for compiling Cython modules.
In order to use a newer version,
please follow the pip instructions for installing from version control
or from a local source tree.
pyunicorn is implemented in Python 3 /
Cython 3, is tested on Linux, macOS
and Windows, and relies on the following packages:
- Required:
- numpy, scipy
- python-igraph (for
Network) - h5netcdf (for
Data,NetCDFDictionary) - tqdm (for progress bars)
- Optional:
- Matplotlib, Cartopy (for plotting features)
- mpi4py (for parallelizing costly computations)
- Sphinx (for generating documentation)
- Jupyter Notebook (for tutorial notebooks)
For extensive HTML documentation, jump right to the homepage. In a local source tree,
HTML and PDF documentation can be generated using Sphinx:
$> pip install .[docs] $> cd docs; make clean html latexpdf
For some example applications look into the tutorials provided with the documentation. They are designed to be self-explanatory, and are set up as Jupyter notebooks.
Before committing changes or opening a pull request (PR) to the code base, please make sure that all tests pass. The test suite is managed by tox and is configured to use system-wide packages when available. Install the test dependencies as follows:
$> pip install -e .[tests]
The test suite can be run from anywhere in the project tree by issuing:
$> tox
To display the defined test environments and target them individually:
$> tox -l $> tox -e style,lint,test,docs
To test individual files:
$> flake8 src/pyunicorn/core/network.py # style check $> pylint src/pyunicorn/core/network.py # static code analysis $> pytest tests/test_core/test_network.py # unit tests