jlizier/jidt

JIDT: Java Information Dynamics Toolkit for studying information-theoretic measures of computation in complex systems

What it solves

Complex systems like brains, swarms, and networks exhibit distributed computation where information is stored, transferred, and modified. JIDT gives researchers reliable, well-tested implementations of information-theoretic measures to quantify these processes from time-series data.

How it works

JIDT computes measures such as transfer entropy, mutual information, active information storage, and conditional variants. It handles both discrete and continuous-valued data. For continuous data, it offers several estimators: Kraskov-Stögbauer-Grassberger (KSG), box-kernel, and linear-Gaussian. All measures are implemented in Java, with wrappers for other languages. A GUI called AutoAnalyser allows point-and-click analysis and even generates code templates.

Who it's for

Researchers studying complex systems—e.g., neuroscience, swarm intelligence, climate, or finance—who want to measure information flow and storage in their data. It's also useful for educators, since JIDT ships with video lectures and tutorials.

Highlights

  • Supports transfer entropy, mutual information, conditional variants, active information storage, and entropy
  • Works with both discrete and continuous data, using estimators like KSG, box-kernel, and linear-Gaussian
  • Provides a GUI (AutoAnalyser) for easy analysis and code generation
  • Can be used from Java, Matlab/Octave, Python, R, Julia, and Clojure
  • Includes demos and a course with short video lectures

Related

  • Project
  • Project
  • Project
  • Project