ennemi 1.5.0

Non-linear correlation detection with mutual information

This package performs non-linear correlation analysis with mutual information (MI). MI is an information-theoretical measure of dependency between two variables. The package is designed for practical data analysis with no theoretical background required.

Features:

• Non-linear correlation detection:
  • Mutual information between two variables, continous or discrete
  • Conditional MI with arbitrary-dimensional conditioning variables
  • Quick overview of many-variable datasets with pairwise MI estimation
• Practical data analysis:
  • Interfaces for evaluating multiple variable pairs and time lags with one call
  • Integrated with pandas data frames (optional)
  • Optimized and automatically parallelized estimation
  • Algorithms verified to work, so that you can focus on your data

This package depends only on NumPy and SciPy; Pandas (2.x or newer) is suggested for more enjoyable data analysis. Recent versions of NumPy 1.x and 2.x are supported. Python 3.11+ on the latest macOS, Ubuntu and Windows versions is officially supported. Older ennemi versions have generally identical behavior if you need to run on older Python.

For more information on theoretical background and usage, please see the documentation. If you encounter any problems or have suggestions, please file an issue!

---

This package was initially developed at Institute for Atmospheric and Earth System Research (INAR), University of Helsinki.