sidpy: Python Tools for Specific Information Dynamics

What is sidpy?

sidpy is a Python package for estimating quantities from specific information dynamics such as specific entropy rate and specific transfer entropy from data, as well as local information dynamics via the Java Information Dynamics Toolbox (JIDT). sidpy also incorporates model selection routines to determine the appropriate model order for scalar and input-output time series.

Installation

pip install git+https://github.com/ddarmon/sidpy

Installation on ARM-based Macs

flann

If you run into an error message like the following after running import sidpy:

File /path/site-packages/pyflann/bindings/flann_ctypes.py:173
    171 flannlib = load_flann_library()
    172 if flannlib == None:
--> 173     raise ImportError('Cannot load dynamic library. Did you compile FLANN?')
    176 class FlannLib:
    177     pass

ImportError: Cannot load dynamic library. Did you compile FLANN?

this is likely because the dynamic library libflann.dylib is not in the pyflann/bindings folder.

To fix this:

1. Install flann

   brew install flann
   

2. Find Installation Location Find where the flann library is installed by running:

   brew info flann
   

   Go to the location provided by brew. On my current installation and version of flann, it is:

   cd /opt/homebrew/Cellar/flann/1.9.2_1
   

3. Copy the libflann.v.v.v.dylib File Finally, copy the libflann.v.v.v.dylib file to site-packages/pyflann as libflann.dylib using:

   mkdir -p /path/site-packages/pyflann/bindings/lib/darwin/
   cp /opt/homebrew/Cellar/flann/1.9.2_1/lib/libflann.1.9.2.dylib /path/site-packages/pyflann/bindings/lib/darwin/libflann.dylib
   

Make sure to replace /path/ with the actual path where your site-packages/pyflann is located. Also, the version number in the libflann.v.v.v.dylib should match the installed version of flann. In this example, it's 1.9.2.

nlopt

If nlopt fails to install using pip, try conda:

conda install nlopt

Using sidpy

See the Jupyter notebooks in example-notebooks for a demonstration of the functionality of sidpy.

If you use sidpy in a scientific publication, please cite this GitHub repository.