An R project to work with entropic coordinates, entropy triangles, NIT and EMA as defined in:
-
(2010) Two information-theoretic tools to assess the performance of multi-class classifiers
-
(2017) The Evaluation of Data Sources using Multivariate Entropy Tools
Uses packages ggtern
and entropy
to provide basic functionality.
Forever in beta testing, since this is the main tool to help me develop the theory, for the time being. At present you can:
-
measure the performance of supervised classification on multiclassifiers (finished. Papers: 2010, 2014, 2020). Also applies to binary channels.
-
measure the information content of multivariate sources (Discrete sources finished. In progress to improve measurements on continuous sources. Paper: 2017)
-
explore the relationship between multilabels in a Multilabel Classification task (Discrete sources. Paper: 2024).
-
measure the information transmission in multivariate data transformations (Still experimental. Paper: 2018)
You may find the theoretical framework, latest news and further support on the project overview:
You can see here some use cases in this link, where you can find some examples of real case scenarios.
This package was developed by me as the original tool (e.g used in the 2010 paper)
Matlab Entropy Triangle Package
The following packages have been developed by students of mine with my approval
There is a plug-in for Weka that lets you evaluate binary classifiers with the ET.
A Python package to work with entropic coordinates and the entropy triangles.
Contact me and I will tell you other initiatives that I am aware of.