/sampling_based_mutual_information

Implements deep-learning and sampling based estimation of mutual information and conditional mutual information

Primary LanguageJupyter NotebookMIT LicenseMIT

This repository implements deep-learning and sampling based methods to estimate the mutual information and conditional mutual information, based on samples (e.g. time points of high-dimensional time-series data).

Please see the example notebook (Sampling_Based_Mutual_Information_Demo) for further explanation and instructions, how to use the classes implemented in sampling_based_mutual_information.py.

Requirements: PyTorch, tqdm, scipy, numpy, matplotlib.