Approximating deficiency-based PIDs for multivariate Gaussian systems
generate.py - Functionality for generating multivariate Gaussian systems (either randomly generated by a Wishart distribution or manually specified) and returning the relevant gain and covariance matrices for determining deficiencies.
estimate.py - Used for computing mutual informations, estimating deficiencies, and computing an approximate deficiency-based partial information decomposion.
experiment.py - Repeatedly sample random systems, approximate the PID, and store the results in results.csv. Running python experiment.py will perform 80,000 simulations, save the results, and then run the hard-coded test cases specified in generate.py.
gen_figure.ipynb - Jupyter notebook for processing the results and generating a figure.