JIDT toolbox is a swiss-army knife of information dynamics analysis of complex systems, with special focus on non-parametric estimators for continuous data. The toolbox can be used from a broad array of modern languages (e.g. Python, Octave/Matlab, R) without the need to install or compile, implements a large variety of measures, and includes GPU support.
The Python package
dit (for Discrete Information Theory) is light and powerful tool to study information-theoretic quantities in discrete-valued systems. The package, which is easily
pip-installable, features a long list of available measures, including a frequently updated suite of partial information decomposition tools.
IDTxl is a Python package specialised in network inference through multivariate transfer entropy. It has built-in data loaders for most common neuroscientific data formats, and supports parallelisation. It serves a purpose similar to
tigramite (see below), although it places more emphasis on non-parametric estimators of multivariate transfer entropy.
tigramite is a Python package devoted to statistical causal discovery for complex systems, extending beyond transfer entropy and with multiple options for modelling and significance testing.
TRENTOOL was one of the first toolboxes for large-scale transfer entropy analysis applied to the neurosciences. It includes a comprehensive set of configurable parameters and advanced statistical significance testing functionality.