Usage of core estimators

IDTxl allows access to core estimators to estimate basic information-theoretic quantities from data. Core-estimators provide no additional functionality like statistical testing or automatic embedding (which are part of the algorithms for network inference or the analysis of network dynamics).

IDTxl provides core estimators for mutual information (MI), conditional mutual information (CMI), transfer entropy (TE), active information storage (AIS), and partial information decomposition (PID). The toolbox provides MI-, CMI-, TE-, and AIS-Estimators for discrete data, jointly Gaussian continuous data, and non-linear continuous data. PID estimators are only available for discrete data.

IDTxl provides two implementations for MI- and CMI-estimators for non-linear continuous data: a multithreaded Java-implementation (JidtKraskovCMI and JidtKraskovMI) and an OpenCL-implementation for the use with GPUs (OpenCLKraskovMI, OpenCLKraskovCMI). See the Installation and Requirements page for software requirements to use either of the two estimator classes.

The core estimators demo script lists all available core estimators and their basic usage (as of August 2018).