MIToolbox

v3.0.1 for C/C++ and MATLAB/Octave

MIToolbox contains a set of functions to calculate information theoreticquantities from data, such as the entropy and mutual information. The toolboxcontains implementations of the most popular Shannon entropies, and also thelesser known Renyi entropy. The toolbox also provides implementations ofthe weighted entropy and weighted mutual information from "Information Theorywith Application", S. Guiasu (1977). The toolbox only supports discrete distributions,as opposed to continuous. All real-valued numbers will be processed by x = floor(x).

These functions are targeted for use with feature selection algorithms ratherthan communication channels and so expect all the data to be available beforeexecution and sample their own probability distributions from the data.

All functions expect the inputs to be vectors or matrices of doubles.

Functions contained:

• Entropy
• Conditional Entropy
• Mutual Information
• Conditional Mutual Information
• generating a joint variable
• generating a probability distribution from a discrete random variable
• Renyi's Entropy
• Renyi's Mutual Information
• Weighted Entropy
• Weighted Mutual Information
• Weighted Conditional Mutual Information

Note: all functions are calculated in log base 2, so return units of "bits".

MIToolbox works on discrete inputs, and all continuous values must bediscretised before use with MIToolbox. Real-valued inputs will be discretisedwith x = floor(x) to ensure compatibility. MIToolbox produces unreliableresults when used with continuous inputs, runs slowly and uses much more memorythan usual. The discrete inputs should have small cardinality, MIToolbox willtreat values {1,10,100} the same way it treats {1,2,3} and the latter will beboth faster and use less memory. This limitation is due to the difficulties inestimating information theoretic functions of continuous variables.

======

Examples:

>> y = [1 1 1 0 0]';
>> x = [1 0 1 1 0]';

>> mi(x,y)       %% mutual information I(X;Y)
ans =
    0.0200

>> h(x)          %% entropy H(X)
ans =
    0.9710

>> condh(x,y)    %% conditional entropy H(X|Y)
ans =
    0.9510

>> h( [x,y] )    %% joint entropy H(X,Y)
ans =
    1.9219

>> joint([x,y])  %% joint random variable XY
ans =
     1
     2
     1
     3
     4

======

All code is licensed under the 3-clause BSD license.

Compilation instructions:

• MATLAB/OCTAVE
  • run CompileMIToolbox.m from the matlab directory.
• Linux C shared library
  • run make x86 or make x64 for 32-bit or 64-bit versions respectively.
  • run sudo make install to install MIToolbox into /usr/local/lib and /usr/local/include
• Windows C dll
  • install MinGW from https://sourceforge.net/projects/mingw-w64/
  • add MinGW binaries folders to PATH, e.g. mingw/bin, mingw/msys/bin
  • run make x64_win to compile a 64-bit Windows dll.

Update History

• 08/02/2017 - v3.0.1 - Bug fix to ensure ANSI C compatibility.
• 07/01/2017 - v3.0.0 - Refactored internals to expose integer information theoretic calculations.
• 10/01/2016 - v2.1.2 - Relicense from LGPL to BSD. Added checks to ensure input MATLAB types are doubles.
• 02/02/2015 - v2.1.1 - Fixed up the Makefile so it installs the headers too.
• 22/02/2014 - v2.1 - Fixed a couple of bugs related to memory handling.Added a make install for compatibility with PyFeast.
• 30/07/2011 - v2.00 - Added implementations of the weighted entropy and weightedmutual information. More cleanup of Mex entry pointto further check the inputs.
• 08/11/2011 - v1.03 - Minor documentation changes to accompany the JMLR publication.
• 15/10/2010 - v1.02 - Fixed bug where MIToolbox would cause a segmentation faultif a x by 0 empty matrix was passed in. Now prints anerror message and returns gracefully.
• 02/09/2010 - v1.01 - Fixed a bug in CMIM.m where the last feature would not beselected first if it had the highest MI.
• 07/07/2010 - v1.00 - Initial Release.