diff options
57 files changed, 6451 insertions, 0 deletions
diff --git a/FEAST/FEAST/FSToolbox/BetaGamma.c b/FEAST/FEAST/FSToolbox/BetaGamma.c new file mode 100644 index 0000000..d500752 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/BetaGamma.c @@ -0,0 +1,187 @@ +/******************************************************************************* +** betaGamma() implements the Beta-Gamma space from Brown (2009). +** This incoporates MIFS, CIFE, and CondRed. +** +** MIFS - "Using mutual information for selecting features in supervised neural net learning" +** R. Battiti, IEEE Transactions on Neural Networks, 1994 +** +** CIFE - "Conditional Infomax Learning: An Integrated Framework for Feature Extraction and Fusion" +** D. Lin and X. Tang, European Conference on Computer Vision (2006) +** +** The Beta Gamma space is explained in Brown (2009) and Brown et al. (2011) +** +** Initial Version - 13/06/2008 +** Updated - 23/06/2011 +** +** Author - Adam Pocock +** +** Part of the Feature Selection Toolbox, please reference +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +** Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +** +** Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +** +** Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +** All rights reserved. +** +** Redistribution and use in source and binary forms, with or without modification, +** are permitted provided that the following conditions are met: +** +** - Redistributions of source code must retain the above copyright notice, this +** list of conditions and the following disclaimer. +** - Redistributions in binary form must reproduce the above copyright notice, +** this list of conditions and the following disclaimer in the documentation +** and/or other materials provided with the distribution. +** - Neither the name of The University of Manchester nor the names of its +** contributors may be used to endorse or promote products derived from this +** software without specific prior written permission. +** +** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +** ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +** WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +** DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +** ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +** (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +** LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +** ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +** SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +** +*******************************************************************************/ + +#include "FSAlgorithms.h" +#include "FSToolbox.h" + +/* MIToolbox includes */ +#include "MutualInformation.h" + +void BetaGamma(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures, double betaParam, double gammaParam) +{ + double **feature2D = (double **) CALLOC_FUNC(noOfFeatures,sizeof(double *)); + + /*holds the class MI values*/ + double *classMI = (double *)CALLOC_FUNC(noOfFeatures,sizeof(double)); + char *selectedFeatures = (char *)CALLOC_FUNC(noOfFeatures,sizeof(char)); + + /*holds the intra feature MI values*/ + int sizeOfMatrix = k*noOfFeatures; + double *featureMIMatrix = (double *)CALLOC_FUNC(sizeOfMatrix,sizeof(double)); + + double maxMI = 0.0; + int maxMICounter = -1; + + double score, currentScore, totalFeatureMI; + int currentHighestFeature, arrayPosition; + + int i,j,m; + + /*********************************************************** + ** because the array is passed as + ** s a m p l e s + ** f + ** e + ** a + ** t + ** u + ** r + ** e + ** s + ** + ** this pulls out a pointer to the first sample of + ** each feature and stores it as a multidimensional array + ** so it can be indexed nicely + ***********************************************************/ + for(j = 0; j < noOfFeatures; j++) + { + feature2D[j] = featureMatrix + (int)j*noOfSamples; + } + + for (i = 0; i < sizeOfMatrix; i++) + { + featureMIMatrix[i] = -1; + }/*for featureMIMatrix - blank to -1*/ + + /*********************************************************** + ** SETUP COMPLETE + ** Algorithm starts here + ***********************************************************/ + + for (i = 0; i < noOfFeatures; i++) + { + classMI[i] = calculateMutualInformation(feature2D[i], classColumn, noOfSamples); + + if (classMI[i] > maxMI) + { + maxMI = classMI[i]; + maxMICounter = i; + }/*if bigger than current maximum*/ + }/*for noOfFeatures - filling classMI*/ + + selectedFeatures[maxMICounter] = 1; + outputFeatures[0] = maxMICounter; + + /************* + ** Now we have populated the classMI array, and selected the highest + ** MI feature as the first output feature + ** Now we move into the JMI algorithm + *************/ + + for (i = 1; i < k; i++) + { + /************************************************************ + ** to ensure it selects some features + ** if this is zero then it will not pick features where the + ** redundancy is greater than the relevance + ************************************************************/ + score = -HUGE_VAL; + currentHighestFeature = 0; + currentScore = 0.0; + totalFeatureMI = 0.0; + + for (j = 0; j < noOfFeatures; j++) + { + /*if we haven't selected j*/ + if (!selectedFeatures[j]) + { + currentScore = classMI[j]; + totalFeatureMI = 0.0; + + for (m = 0; m < i; m++) + { + arrayPosition = m*noOfFeatures + j; + if (featureMIMatrix[arrayPosition] == -1) + { + /*double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength);*/ + featureMIMatrix[arrayPosition] = betaParam*calculateMutualInformation(feature2D[(int) outputFeatures[m]], feature2D[j], noOfSamples); + + /*double calculateConditionalMutualInformation(double *firstVector, double *targetVector, double* conditionVector, int vectorLength);*/ + featureMIMatrix[arrayPosition] -= gammaParam*calculateConditionalMutualInformation(feature2D[(int) outputFeatures[m]], feature2D[j], classColumn, noOfSamples); + }/*if not already known*/ + + totalFeatureMI += featureMIMatrix[arrayPosition]; + }/*for the number of already selected features*/ + + currentScore -= (totalFeatureMI); + + if (currentScore > score) + { + score = currentScore; + currentHighestFeature = j; + } + }/*if j is unselected*/ + }/*for number of features*/ + + selectedFeatures[currentHighestFeature] = 1; + outputFeatures[i] = currentHighestFeature; + + }/*for the number of features to select*/ + + for (i = 0; i < k; i++) + { + outputFeatures[i] += 1; /*C++ indexes from 0 not 1*/ + }/*for number of selected features*/ + +}/*BetaGamma(int,int,int,double[][],double[],double[],double,double)*/ + diff --git a/FEAST/FEAST/FSToolbox/CMIM.c b/FEAST/FEAST/FSToolbox/CMIM.c new file mode 100644 index 0000000..8815644 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/CMIM.c @@ -0,0 +1,141 @@ +/******************************************************************************* +** CMIM.c, implements a discrete version of the +** Conditional Mutual Information Maximisation criterion, using the fast +** exact implementation from +** +** "Fast Binary Feature Selection using Conditional Mutual Information Maximisation" +** F. Fleuret, JMLR (2004) +** +** Initial Version - 13/06/2008 +** Updated - 23/06/2011 +** +** Author - Adam Pocock +** +** Part of the Feature Selection Toolbox, please reference +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +** Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +** +** Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +** +** Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +** All rights reserved. +** +** Redistribution and use in source and binary forms, with or without modification, +** are permitted provided that the following conditions are met: +** +** - Redistributions of source code must retain the above copyright notice, this +** list of conditions and the following disclaimer. +** - Redistributions in binary form must reproduce the above copyright notice, +** this list of conditions and the following disclaimer in the documentation +** and/or other materials provided with the distribution. +** - Neither the name of The University of Manchester nor the names of its +** contributors may be used to endorse or promote products derived from this +** software without specific prior written permission. +** +** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +** ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +** WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +** DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +** ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +** (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +** LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +** ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +** SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +** +*******************************************************************************/ + +#include "FSAlgorithms.h" +#include "FSToolbox.h" + +/* MIToolbox includes */ +#include "MutualInformation.h" + +void CMIM(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures) +{ + /*holds the class MI values + **the class MI doubles as the partial score from the CMIM paper + */ + double *classMI = (double *)CALLOC_FUNC(noOfFeatures,sizeof(double)); + /*in the CMIM paper, m = lastUsedFeature*/ + int *lastUsedFeature = (int *)CALLOC_FUNC(noOfFeatures,sizeof(int)); + + double score, conditionalInfo; + int iMinus, currentFeature; + + double maxMI = 0.0; + int maxMICounter = -1; + + int j,i; + + double **feature2D = (double**) CALLOC_FUNC(noOfFeatures,sizeof(double*)); + + for(j = 0; j < noOfFeatures; j++) + { + feature2D[j] = featureMatrix + (int)j*noOfSamples; + } + + for (i = 0; i < noOfFeatures;i++) + { + classMI[i] = calculateMutualInformation(feature2D[i], classColumn, noOfSamples); + + if (classMI[i] > maxMI) + { + maxMI = classMI[i]; + maxMICounter = i; + }/*if bigger than current maximum*/ + }/*for noOfFeatures - filling classMI*/ + + outputFeatures[0] = maxMICounter; + + /***************************************************************************** + ** We have populated the classMI array, and selected the highest + ** MI feature as the first output feature + ** Now we move into the CMIM algorithm + *****************************************************************************/ + + for (i = 1; i < k; i++) + { + score = 0.0; + iMinus = i-1; + + for (j = 0; j < noOfFeatures; j++) + { + while ((classMI[j] > score) && (lastUsedFeature[j] < i)) + { + /*double calculateConditionalMutualInformation(double *firstVector, double *targetVector, double *conditionVector, int vectorLength);*/ + currentFeature = (int) outputFeatures[lastUsedFeature[j]]; + conditionalInfo = calculateConditionalMutualInformation(feature2D[j],classColumn,feature2D[currentFeature],noOfSamples); + if (classMI[j] > conditionalInfo) + { + classMI[j] = conditionalInfo; + }/*reset classMI*/ + /*moved due to C indexing from 0 rather than 1*/ + lastUsedFeature[j] += 1; + }/*while partial score greater than score & not reached last feature*/ + if (classMI[j] > score) + { + score = classMI[j]; + outputFeatures[i] = j; + }/*if partial score still greater than score*/ + }/*for number of features*/ + }/*for the number of features to select*/ + + + for (i = 0; i < k; i++) + { + outputFeatures[i] += 1; /*C indexes from 0 not 1*/ + }/*for number of selected features*/ + + FREE_FUNC(classMI); + FREE_FUNC(lastUsedFeature); + FREE_FUNC(feature2D); + + classMI = NULL; + lastUsedFeature = NULL; + feature2D = NULL; + +}/*CMIM(int,int,int,double[][],double[],double[])*/ + diff --git a/FEAST/FEAST/FSToolbox/CompileFEAST.m b/FEAST/FEAST/FSToolbox/CompileFEAST.m new file mode 100644 index 0000000..f5dad48 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/CompileFEAST.m @@ -0,0 +1,4 @@ +%Compiles the FEAST Toolbox into a mex executable for use with MATLAB + +mex -I../MIToolbox FSToolboxMex.c BetaGamma.c CMIM.c CondMI.c DISR.c ICAP.c JMI.c mRMR_D.c ../MIToolbox/MutualInformation.c ../MIToolbox/Entropy.c ../MIToolbox/CalculateProbability.c ../MIToolbox/ArrayOperations.c + diff --git a/FEAST/FEAST/FSToolbox/CondMI.c b/FEAST/FEAST/FSToolbox/CondMI.c new file mode 100644 index 0000000..e1f746a --- /dev/null +++ b/FEAST/FEAST/FSToolbox/CondMI.c @@ -0,0 +1,164 @@ +/******************************************************************************* +** CondMI.c, implements the CMI criterion using a greedy forward search +** +** Initial Version - 19/08/2010 +** Updated - 23/06/2011 +** +** Author - Adam Pocock +** +** Part of the Feature Selection Toolbox, please reference +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +** Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +** +** Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +** +** Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +** All rights reserved. +** +** Redistribution and use in source and binary forms, with or without modification, +** are permitted provided that the following conditions are met: +** +** - Redistributions of source code must retain the above copyright notice, this +** list of conditions and the following disclaimer. +** - Redistributions in binary form must reproduce the above copyright notice, +** this list of conditions and the following disclaimer in the documentation +** and/or other materials provided with the distribution. +** - Neither the name of The University of Manchester nor the names of its +** contributors may be used to endorse or promote products derived from this +** software without specific prior written permission. +** +** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +** ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +** WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +** DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +** ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +** (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +** LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +** ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +** SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +** +*******************************************************************************/ + +#include "FSAlgorithms.h" +#include "FSToolbox.h" + +/* for memcpy */ +#include <string.h> + +/* MIToolbox includes */ +#include "MutualInformation.h" +#include "ArrayOperations.h" + +void CondMI(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures) +{ + /*holds the class MI values*/ + double *classMI = (double *)CALLOC_FUNC(noOfFeatures,sizeof(double)); + + char *selectedFeatures = (char *)CALLOC_FUNC(noOfFeatures,sizeof(char)); + + /*holds the intra feature MI values*/ + int sizeOfMatrix = k*noOfFeatures; + double *featureMIMatrix = (double *)CALLOC_FUNC(sizeOfMatrix,sizeof(double)); + + double maxMI = 0.0; + int maxMICounter = -1; + + double **feature2D = (double**) CALLOC_FUNC(noOfFeatures,sizeof(double*)); + + double score, currentScore, totalFeatureMI; + int currentHighestFeature; + + double *conditionVector = (double *) CALLOC_FUNC(noOfSamples,sizeof(double)); + + int arrayPosition; + double mi, tripEntropy; + + int i,j,x; + + for(j = 0; j < noOfFeatures; j++) + { + feature2D[j] = featureMatrix + (int)j*noOfSamples; + } + + for (i = 0; i < sizeOfMatrix;i++) + { + featureMIMatrix[i] = -1; + }/*for featureMIMatrix - blank to -1*/ + + for (i = 0; i < noOfFeatures;i++) + { + /*calculate mutual info + **double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength); + */ + classMI[i] = calculateMutualInformation(feature2D[i], classColumn, noOfSamples); + + if (classMI[i] > maxMI) + { + maxMI = classMI[i]; + maxMICounter = i; + }/*if bigger than current maximum*/ + }/*for noOfFeatures - filling classMI*/ + + selectedFeatures[maxMICounter] = 1; + outputFeatures[0] = maxMICounter; + + memcpy(conditionVector,feature2D[maxMICounter],sizeof(double)*noOfSamples); + + /***************************************************************************** + ** We have populated the classMI array, and selected the highest + ** MI feature as the first output feature + ** Now we move into the CondMI algorithm + *****************************************************************************/ + + for (i = 1; i < k; i++) + { + score = 0.0; + currentHighestFeature = -1; + currentScore = 0.0; + totalFeatureMI = 0.0; + + for (j = 0; j < noOfFeatures; j++) + { + /*if we haven't selected j*/ + if (selectedFeatures[j] == 0) + { + currentScore = 0.0; + totalFeatureMI = 0.0; + + /*double calculateConditionalMutualInformation(double *firstVector, double *targetVector, double *conditionVector, int vectorLength);*/ + currentScore = calculateConditionalMutualInformation(feature2D[j],classColumn,conditionVector,noOfSamples); + + if (currentScore > score) + { + score = currentScore; + currentHighestFeature = j; + } + }/*if j is unselected*/ + }/*for number of features*/ + + outputFeatures[i] = currentHighestFeature; + + if (currentHighestFeature != -1) + { + selectedFeatures[currentHighestFeature] = 1; + mergeArrays(feature2D[currentHighestFeature],conditionVector,conditionVector,noOfSamples); + } + + }/*for the number of features to select*/ + + FREE_FUNC(classMI); + FREE_FUNC(conditionVector); + FREE_FUNC(feature2D); + FREE_FUNC(featureMIMatrix); + FREE_FUNC(selectedFeatures); + + classMI = NULL; + conditionVector = NULL; + feature2D = NULL; + featureMIMatrix = NULL; + selectedFeatures = NULL; +}/*CondMI(int,int,int,double[][],double[],double[])*/ + diff --git a/FEAST/FEAST/FSToolbox/DISR.c b/FEAST/FEAST/FSToolbox/DISR.c new file mode 100644 index 0000000..f8b19e6 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/DISR.c @@ -0,0 +1,181 @@ +/******************************************************************************* +** DISR.c, implements the Double Input Symmetrical Relevance criterion +** from +** +** "On the Use of Variable Complementarity for Feature Selection in Cancer Classification" +** P. Meyer and G. Bontempi, (2006) +** +** Initial Version - 13/06/2008 +** Updated - 23/06/2011 +** +** Author - Adam Pocock +** +** Part of the Feature Selection Toolbox, please reference +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +** Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +** +** Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +** +** Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +** All rights reserved. +** +** Redistribution and use in source and binary forms, with or without modification, +** are permitted provided that the following conditions are met: +** +** - Redistributions of source code must retain the above copyright notice, this +** list of conditions and the following disclaimer. +** - Redistributions in binary form must reproduce the above copyright notice, +** this list of conditions and the following disclaimer in the documentation +** and/or other materials provided with the distribution. +** - Neither the name of The University of Manchester nor the names of its +** contributors may be used to endorse or promote products derived from this +** software without specific prior written permission. +** +** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +** ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +** WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +** DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +** ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +** (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +** LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +** ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +** SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +** +*******************************************************************************/ +#include "FSAlgorithms.h" +#include "FSToolbox.h" + +/* MIToolbox includes */ +#include "MutualInformation.h" +#include "Entropy.h" +#include "ArrayOperations.h" + +void DISR(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures) +{ + /*holds the class MI values*/ + double *classMI = (double *)CALLOC_FUNC(noOfFeatures,sizeof(double)); + + char *selectedFeatures = (char *)CALLOC_FUNC(noOfFeatures,sizeof(char)); + + /*holds the intra feature MI values*/ + int sizeOfMatrix = k*noOfFeatures; + double *featureMIMatrix = (double *)CALLOC_FUNC(sizeOfMatrix,sizeof(double)); + + double maxMI = 0.0; + int maxMICounter = -1; + + double **feature2D = (double**) CALLOC_FUNC(noOfFeatures,sizeof(double*)); + + double score, currentScore, totalFeatureMI; + int currentHighestFeature; + + double *mergedVector = (double *) CALLOC_FUNC(noOfSamples,sizeof(double)); + + int arrayPosition; + double mi, tripEntropy; + + int i,j,x; + + for(j = 0; j < noOfFeatures; j++) + { + feature2D[j] = featureMatrix + (int)j*noOfSamples; + } + + for (i = 0; i < sizeOfMatrix;i++) + { + featureMIMatrix[i] = -1; + }/*for featureMIMatrix - blank to -1*/ + + + for (i = 0; i < noOfFeatures;i++) + { + /*calculate mutual info + **double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength); + */ + classMI[i] = calculateMutualInformation(feature2D[i], classColumn, noOfSamples); + + if (classMI[i] > maxMI) + { + maxMI = classMI[i]; + maxMICounter = i; + }/*if bigger than current maximum*/ + }/*for noOfFeatures - filling classMI*/ + + selectedFeatures[maxMICounter] = 1; + outputFeatures[0] = maxMICounter; + + /***************************************************************************** + ** We have populated the classMI array, and selected the highest + ** MI feature as the first output feature + ** Now we move into the DISR algorithm + *****************************************************************************/ + + for (i = 1; i < k; i++) + { + score = 0.0; + currentHighestFeature = 0; + currentScore = 0.0; + totalFeatureMI = 0.0; + + for (j = 0; j < noOfFeatures; j++) + { + /*if we haven't selected j*/ + if (selectedFeatures[j] == 0) + { + currentScore = 0.0; + totalFeatureMI = 0.0; + + for (x = 0; x < i; x++) + { + arrayPosition = x*noOfFeatures + j; + if (featureMIMatrix[arrayPosition] == -1) + { + /* + **double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength); + **double calculateJointEntropy(double *firstVector, double *secondVector, int vectorLength); + */ + + mergeArrays(feature2D[(int) outputFeatures[x]], feature2D[j],mergedVector,noOfSamples); + mi = calculateMutualInformation(mergedVector, classColumn, noOfSamples); + tripEntropy = calculateJointEntropy(mergedVector, classColumn, noOfSamples); + + featureMIMatrix[arrayPosition] = mi / tripEntropy; + }/*if not already known*/ + currentScore += featureMIMatrix[arrayPosition]; + }/*for the number of already selected features*/ + + if (currentScore > score) + { + score = currentScore; + currentHighestFeature = j; + } + }/*if j is unselected*/ + }/*for number of features*/ + + selectedFeatures[currentHighestFeature] = 1; + outputFeatures[i] = currentHighestFeature; + + }/*for the number of features to select*/ + + for (i = 0; i < k; i++) + { + outputFeatures[i] += 1; /*C indexes from 0 not 1*/ + }/*for number of selected features*/ + + FREE_FUNC(classMI); + FREE_FUNC(mergedVector); + FREE_FUNC(feature2D); + FREE_FUNC(featureMIMatrix); + FREE_FUNC(selectedFeatures); + + classMI = NULL; + mergedVector = NULL; + feature2D = NULL; + featureMIMatrix = NULL; + selectedFeatures = NULL; + +}/*DISR(int,int,int,double[][],double[],double[])*/ + diff --git a/FEAST/FEAST/FSToolbox/FCBF.m b/FEAST/FEAST/FSToolbox/FCBF.m new file mode 100644 index 0000000..dcaf3bf --- /dev/null +++ b/FEAST/FEAST/FSToolbox/FCBF.m @@ -0,0 +1,58 @@ +function [selectedFeatures] = FCBF(featureMatrix,classColumn,threshold) +%function [selectedFeatures] = FCBF(featureMatrix,classColumn,threshold) +% +%Performs feature selection using the FCBF measure by Yu and Liu 2004. +% +%Instead of selecting a fixed number of features it provides a relevancy threshold and selects all +%features which score above that and are not redundant +% +% The license is in the license.txt provided. + +numFeatures = size(featureMatrix,2); +classScore = zeros(numFeatures,1); + +for i = 1:numFeatures + classScore(i) = SU(featureMatrix(:,i),classColumn); +end + +[classScore indexScore] = sort(classScore,1,'descend'); + +indexScore = indexScore(classScore > threshold); +classScore = classScore(classScore > threshold); + +if ~isempty(indexScore) + curPosition = 1; +else + curPosition = 0; +end + +while curPosition <= length(indexScore) + j = curPosition + 1; + curFeature = indexScore(curPosition); + while j <= length(indexScore) + scoreij = SU(featureMatrix(:,curFeature),featureMatrix(:,indexScore(j))); + if scoreij > classScore(j) + indexScore(j) = []; + classScore(j) = []; + else + j = j + 1; + end + end + curPosition = curPosition + 1; +end + +selectedFeatures = indexScore; + +end + +function [score] = SU(firstVector,secondVector) +%function [score] = SU(firstVector,secondVector) +% +%calculates SU = 2 * (I(X;Y)/(H(X) + H(Y))) + +hX = h(firstVector); +hY = h(secondVector); +iXY = mi(firstVector,secondVector); + +score = (2 * iXY) / (hX + hY); +end diff --git a/FEAST/FEAST/FSToolbox/FSAlgorithms.h b/FEAST/FEAST/FSToolbox/FSAlgorithms.h new file mode 100644 index 0000000..f53990c --- /dev/null +++ b/FEAST/FEAST/FSToolbox/FSAlgorithms.h @@ -0,0 +1,138 @@ +/******************************************************************************* +** +** FSAlgorithms.h +** Provides the function definitions for the list of algorithms implemented +** in the FSToolbox. +** +** Author: Adam Pocock +** Created: 27/06/2011 +** +** Copyright 2010/2011 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** Part of the FEAture Selection Toolbox (FEAST), please reference +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +** Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +** +** +** Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +** +** Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +** All rights reserved. +** +** Redistribution and use in source and binary forms, with or without modification, +** are permitted provided that the following conditions are met: +** +** - Redistributions of source code must retain the above copyright notice, this +** list of conditions and the following disclaimer. +** - Redistributions in binary form must reproduce the above copyright notice, +** this list of conditions and the following disclaimer in the documentation +** and/or other materials provided with the distribution. +** - Neither the name of The University of Manchester nor the names of its +** contributors may be used to endorse or promote products derived from this +** software without specific prior written permission. +** +** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +** ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +** WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +** DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +** ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +** (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +** LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +** ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +** SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +** +*******************************************************************************/ + +/******************************************************************************* + * All algorithms take an integer k which determines how many features to + * select, the number of samples and the number of features. Additionally each + * algorithm takes pointers to the data matrix, and the label vector, and + * a pointer to the output vector. The output vector should be pre-allocated + * with sizeof(double)*k bytes. + * + * Some algorithms take additional parameters, which given at the end of the + * standard parameter list. + * + * Each algorithm uses a forward search, and selects the feature which has + * the maxmimum MI with the labels first. + * + * All the algorithms except CMIM use an optimised variant which caches the + * previously calculated MI values. This trades space for time, but can + * allocate large amounts of memory. CMIM uses the optimised implementation + * given in Fleuret (2004). + *****************************************************************************/ + +#ifndef __FSAlgorithms_H +#define __FSAlgorithms_H + +/******************************************************************************* +** mRMR_D() implements the minimum Relevance Maximum Redundancy criterion +** using the difference variant, from +** +** "Feature Selection Based on Mutual Information: Criteria of Max-Dependency, Max-Relevance, and Min-Redundancy" +** H. Peng et al. IEEE Pattern Analysis and Machine Intelligence (PAMI) (2005) +*******************************************************************************/ +void mRMR_D(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures); + +/******************************************************************************* +** CMIM() implements a discrete version of the +** Conditional Mutual Information Maximisation criterion, using the fast +** exact implementation from +** +** "Fast Binary Feature Selection using Conditional Mutual Information Maximisation" +** F. Fleuret, JMLR (2004) +*******************************************************************************/ +void CMIM(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures); + +/******************************************************************************* +** JMI() implements the JMI criterion from +** +** "Data Visualization and Feature Selection: New Algorithms for Nongaussian Data" +** H. Yang and J. Moody, NIPS (1999) +*******************************************************************************/ +void JMI(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures); + +/******************************************************************************* +** DISR() implements the Double Input Symmetrical Relevance criterion +** from +** +** "On the Use of Variable Complementarity for Feature Selection in Cancer Classification" +** P. Meyer and G. Bontempi, (2006) +*******************************************************************************/ +void DISR(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures); + +/******************************************************************************* +** ICAP() implements the Interaction Capping criterion from +** +** "Machine Learning Based on Attribute Interactions" +** A. Jakulin, PhD Thesis (2005) +*******************************************************************************/ +void ICAP(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures); + +/******************************************************************************* +** CondMI() implements the CMI criterion using a greedy forward search +*******************************************************************************/ +void CondMI(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures); + +/******************************************************************************* +** betaGamma() implements the Beta-Gamma space from Brown (2009). +** This incoporates MIFS, CIFE, and CondRed. +** +** MIFS - "Using mutual information for selecting features in supervised neural net learning" +** R. Battiti, IEEE Transactions on Neural Networks, 1994 +** +** CIFE - "Conditional Infomax Learning: An Integrated Framework for Feature Extraction and Fusion" +** D. Lin and X. Tang, European Conference on Computer Vision (2006) +** +** The Beta Gamma space is explained in our paper +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +*******************************************************************************/ +void BetaGamma(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures, double beta, double gamma); + +#endif diff --git a/FEAST/FEAST/FSToolbox/FSToolbox.h b/FEAST/FEAST/FSToolbox/FSToolbox.h new file mode 100644 index 0000000..bf8662b --- /dev/null +++ b/FEAST/FEAST/FSToolbox/FSToolbox.h @@ -0,0 +1,70 @@ +/******************************************************************************* ** +** FSToolbox.h +** Provides the header files and #defines to ensure compatibility with MATLAB +** and C/C++. By default it compiles to MATLAB, if COMPILE_C is defined it +** links to the C memory allocation functions. +** +** Author: Adam Pocock +** Created: 27/06/2011 +** +** Copyright 2010/2011 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** Part of the FEAture Selection Toolbox (FEAST), please reference +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +** Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +** +** Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +** +** Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +** All rights reserved. +** +** Redistribution and use in source and binary forms, with or without modification, +** are permitted provided that the following conditions are met: +** +** - Redistributions of source code must retain the above copyright notice, this +** list of conditions and the following disclaimer. +** - Redistributions in binary form must reproduce the above copyright notice, +** this list of conditions and the following disclaimer in the documentation +** and/or other materials provided with the distribution. +** - Neither the name of The University of Manchester nor the names of its +** contributors may be used to endorse or promote products derived from this +** software without specific prior written permission. +** +** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +** ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +** WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +** DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +** ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +** (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +** LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +** ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +** SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +** +*******************************************************************************/ + +#ifndef __FSToolbox_H +#define __FSToolbox_H + +#include <math.h> +#include <string.h> + +#ifdef COMPILE_C + #define C_IMPLEMENTATION + #include <stdio.h> + #include <stdlib.h> + #define CALLOC_FUNC calloc + #define FREE_FUNC free +#else + #define MEX_IMPLEMENTATION + #include "mex.h" + #define CALLOC_FUNC mxCalloc + #define FREE_FUNC mxFree + #define printf mexPrintf /*for Octave-3.2*/ +#endif + +#endif + diff --git a/FEAST/FEAST/FSToolbox/FSToolboxMex.c b/FEAST/FEAST/FSToolbox/FSToolboxMex.c new file mode 100644 index 0000000..73a9197 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/FSToolboxMex.c @@ -0,0 +1,290 @@ +/******************************************************************************* +** FSToolboxMex.c is the entry point for the feature selection toolbox. +** It provides a MATLAB interface to the various selection algorithms. +** +** Initial Version - 27/06/2011 +** +** Author - Adam Pocock +** +** Part of the Feature Selection Toolbox, please reference +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +** Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +** +** Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +** +** Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +** All rights reserved. +** +** Redistribution and use in source and binary forms, with or without modification, +** are permitted provided that the following conditions are met: +** +** - Redistributions of source code must retain the above copyright notice, this +** list of conditions and the following disclaimer. +** - Redistributions in binary form must reproduce the above copyright notice, +** this list of conditions and the following disclaimer in the documentation +** and/or other materials provided with the distribution. +** - Neither the name of The University of Manchester nor the names of its +** contributors may be used to endorse or promote products derived from this +** software without specific prior written permission. +** +** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +** ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +** WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +** DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +** ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +** (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +** LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +** ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +** SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +** +*******************************************************************************/ + +#include "FSToolbox.h" +#include "FSAlgorithms.h" +#include "Entropy.h" + +/****************************************************************************** +** entry point for the mex call +** nlhs - number of outputs +** plhs - pointer to array of outputs +** nrhs - number of inputs +** prhs - pointer to array of inputs +******************************************************************************/ +void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) +{ + /************************************************************* + ** this function takes 4-6 arguments: + ** flag = which algorithm to use, + ** k = number of features to select, + ** featureMatrix[][] = matrix of features, + ** classColumn[] = targets, + ** optionalParam1 = (path angle or beta value), + ** optionalParam2 = (gamma value), + ** the arguments should all be discrete integers. + ** and has one output: + ** selectedFeatures[] of size k + *************************************************************/ + + int flag, k; + double optionalParam1, optionalParam2; + int numberOfFeatures, numberOfSamples, numberOfTargets; + double *featureMatrix, *targets, *output, *outputFeatures; + + double entropyTest; + int i,j; + + /************************************************************ + ** number to function map + ** 1 = MIFS + ** 2 = mRMR + ** 3 = CMIM + ** 4 = JMI + ** 5 = DISR + ** 6 = CIFE + ** 7 = ICAP + ** 8 = CondRed + ** 9 = BetaGamma + ** 10 = CMI + *************************************************************/ + if (nlhs > 1) + { + printf("Incorrect number of output arguments"); + }/*if not 1 output*/ + if ((nrhs < 4) || (nrhs > 6)) + { + printf("Incorrect number of input arguments"); + return; + }/*if not 4-6 inputs*/ + + /*get the flag for which algorithm*/ + flag = (int) mxGetScalar(prhs[0]); + + /*get the number of features to select, cast out as it is a double*/ + k = (int) mxGetScalar(prhs[1]); + + numberOfFeatures = mxGetN(prhs[2]); + numberOfSamples = mxGetM(prhs[2]); + + numberOfTargets = mxGetM(prhs[3]); + + if (nrhs == 6) + { + optionalParam1 = (double) mxGetScalar(prhs[4]); + optionalParam2 = (double) mxGetScalar(prhs[5]); + } + else if (nrhs == 5) + { + optionalParam1 = (double) mxGetScalar(prhs[4]); + optionalParam2 = 0.0; + } + + if (numberOfTargets != numberOfSamples) + { + printf("Number of targets must match number of samples\n"); + printf("Number of targets = %d, Number of Samples = %d, Number of Features = %d\n",numberOfTargets,numberOfSamples,numberOfFeatures); + + plhs[0] = mxCreateDoubleMatrix(0,0,mxREAL); + return; + }/*if size mismatch*/ + else if ((k < 1) || (k > numberOfFeatures)) + { + printf("You have requested k = %d features, which is not possible\n",k); + plhs[0] = mxCreateDoubleMatrix(0,0,mxREAL); + return; + } + else + { + featureMatrix = mxGetPr(prhs[2]); + targets = mxGetPr(prhs[3]); + + /*double calculateEntropy(double *dataVector, int vectorLength)*/ + entropyTest = calculateEntropy(targets,numberOfSamples); + if (entropyTest < 0.0000001) + { + printf("The class label Y has entropy of 0, therefore all mutual informations containing Y will be 0. No feature selection is performed\n"); + plhs[0] = mxCreateDoubleMatrix(0,0,mxREAL); + return; + } + else + { + /*printf("Flag = %d, k = %d, numFeatures = %d, numSamples = %d\n",flag,k,numberOfFeatures,numberOfSamples);*/ + switch (flag) + { + case 1: /* MIFS */ + { + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + if (nrhs == 4) + { + /* MIFS is Beta = 1, Gamma = 0 */ + optionalParam1 = 1.0; + optionalParam2 = 0.0; + } + + /*void BetaGamma(int k, long noOfSamples, long noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures, double beta, double gamma)*/ + BetaGamma(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output,optionalParam1,optionalParam2); + break; + } + case 2: /* mRMR */ + { + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /*void mRMR_D(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures)*/ + mRMR_D(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output); + break; + } + case 3: /* CMIM */ + { + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /*void CMIM(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures)*/ + CMIM(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output); + break; + } + case 4: /* JMI */ + { + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /*void JMI(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures)*/ + JMI(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output); + break; + } + case 5: /* DISR */ + { + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /*void DISR(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures)*/ + DISR(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output); + break; + } + case 6: /* CIFE */ + { + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /* CIFE is Beta = 1, Gamma = 1 */ + optionalParam1 = 1.0; + optionalParam2 = 1.0; + + /*void BetaGamma(int k, long noOfSamples, long noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures, double beta, double gamma)*/ + BetaGamma(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output,optionalParam1,optionalParam2); + break; + } + case 7: /* ICAP */ + { + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /*void ICAP(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output);*/ + ICAP(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output); + break; + } + case 8: /* CondRed */ + { + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /* CondRed is Beta = 0, Gamma = 1 */ + optionalParam1 = 0.0; + optionalParam2 = 1.0; + + /*void BetaGamma(int k, long noOfSamples, long noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures, double beta, double gamma)*/ + BetaGamma(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output,optionalParam1,optionalParam2); + break; + } + case 9: /* BetaGamma */ + { + if (nrhs != 6) + { + printf("Insufficient arguments specified for Beta Gamma FS\n"); + plhs[0] = mxCreateDoubleMatrix(0,0,mxREAL); + return; + } + else + { + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /*void BetaGamma(int k, long noOfSamples, long noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures, double beta, double gamma)*/ + BetaGamma(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output,optionalParam1,optionalParam2); + } + break; + } + case 10: /* CMI */ + { + output = (double *)mxCalloc(k,sizeof(double)); + + /*void CondMI(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures)*/ + CondMI(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output); + + i = 0; + + while((output[i] != -1) && (i < k)) + { + i++; + } + + plhs[0] = mxCreateDoubleMatrix(i,1,mxREAL); + outputFeatures = (double *)mxGetPr(plhs[0]); + + for (j = 0; j < i; j++) + { + outputFeatures[j] = output[j] + 1; /*C indexes from 0 not 1*/ + }/*for number of selected features*/ + + mxFree(output); + output = NULL; + break; + } + }/*switch on flag*/ + return; + } + } +}/*mex function entry*/ diff --git a/FEAST/FEAST/FSToolbox/FSToolboxMex.mexa64 b/FEAST/FEAST/FSToolbox/FSToolboxMex.mexa64 Binary files differnew file mode 100755 index 0000000..0015254 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/FSToolboxMex.mexa64 diff --git a/FEAST/FEAST/FSToolbox/ICAP.c b/FEAST/FEAST/FSToolbox/ICAP.c new file mode 100644 index 0000000..8f1e260 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/ICAP.c @@ -0,0 +1,182 @@ +/******************************************************************************* +** ICAP.c implements the Interaction Capping criterion from +** +** "Machine Learning Based on Attribute Interactions" +** A. Jakulin, PhD Thesis (2005) +** +** Initial Version - 19/08/2010 +** Updated - 23/06/2011 +** +** Author - Adam Pocock +** +** Part of the Feature Selection Toolbox, please reference +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +** Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +** +** Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +** +** Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +** All rights reserved. +** +** Redistribution and use in source and binary forms, with or without modification, +** are permitted provided that the following conditions are met: +** +** - Redistributions of source code must retain the above copyright notice, this +** list of conditions and the following disclaimer. +** - Redistributions in binary form must reproduce the above copyright notice, +** this list of conditions and the following disclaimer in the documentation +** and/or other materials provided with the distribution. +** - Neither the name of The University of Manchester nor the names of its +** contributors may be used to endorse or promote products derived from this +** software without specific prior written permission. +** +** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +** ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +** WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +** DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +** ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +** (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +** LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +** ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +** SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +** +*******************************************************************************/ + +#include "FSAlgorithms.h" +#include "FSToolbox.h" + +/* MIToolbox includes */ +#include "MutualInformation.h" + +void ICAP(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures) +{ + /*holds the class MI values*/ + double *classMI = (double *)CALLOC_FUNC(noOfFeatures,sizeof(double)); + char *selectedFeatures = (char *)CALLOC_FUNC(noOfFeatures,sizeof(char)); + + /*separates out the features*/ + double **feature2D = (double **) CALLOC_FUNC(noOfFeatures,sizeof(double *)); + + /*holds the intra feature MI values*/ + int sizeOfMatrix = k*noOfFeatures; + double *featureMIMatrix = (double *)CALLOC_FUNC(sizeOfMatrix,sizeof(double)); + double *featureCMIMatrix = (double *)CALLOC_FUNC(sizeOfMatrix,sizeof(double)); + + double maxMI = 0.0; + int maxMICounter = -1; + + double score, currentScore, totalFeatureInteraction, interactionInfo; + int currentHighestFeature, arrayPosition; + + int i, j, m; + + for (j = 0; j < noOfFeatures; j++) + feature2D[j] = featureMatrix + (int) j * noOfSamples; + + for (i = 0; i < sizeOfMatrix; i++) + { + featureMIMatrix[i] = -1; + featureCMIMatrix[i] = -1; + }/*for featureMIMatrix and featureCMIMatrix - blank to -1*/ + + /*SETUP COMPLETE*/ + /*Algorithm starts here*/ + + for (i = 0; i < noOfFeatures;i++) + { + classMI[i] = calculateMutualInformation(feature2D[i], classColumn, noOfSamples); + + if (classMI[i] > maxMI) + { + maxMI = classMI[i]; + maxMICounter = i; + }/*if bigger than current maximum*/ + }/*for noOfFeatures - filling classMI*/ + + selectedFeatures[maxMICounter] = 1; + outputFeatures[0] = maxMICounter; + + /************* + ** Now we have populated the classMI array, and selected the highest + ** MI feature as the first output feature + *************/ + + for (i = 1; i < k; i++) + { + /********************************************************************** + ** to ensure it selects some features + **if this is zero then it will not pick features where the redundancy is greater than the + **relevance + **********************************************************************/ + score = -HUGE_VAL; + currentHighestFeature = 0; + currentScore = 0.0; + + for (j = 0; j < noOfFeatures; j++) + { + /*if we haven't selected j*/ + if (!selectedFeatures[j]) + { + currentScore = classMI[j]; + totalFeatureInteraction = 0.0; + + for (m = 0; m < i; m++) + { + arrayPosition = m*noOfFeatures + j; + + if (featureMIMatrix[arrayPosition] == -1) + { + /*work out interaction*/ + + /*double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength);*/ + featureMIMatrix[arrayPosition] = calculateMutualInformation(feature2D[(int) outputFeatures[m]], feature2D[j], noOfSamples); + /*double calculateConditionalMutualInformation(double *firstVector, double *targetVector, double* conditionVector, int vectorLength);*/ + featureCMIMatrix[arrayPosition] = calculateConditionalMutualInformation(feature2D[(int) outputFeatures[m]], feature2D[j], classColumn, noOfSamples); + }/*if not already known*/ + + interactionInfo = featureCMIMatrix[arrayPosition] - featureMIMatrix[arrayPosition]; + + if (interactionInfo < 0) + { + totalFeatureInteraction += interactionInfo; + } + }/*for the number of already selected features*/ + + currentScore += totalFeatureInteraction; + + + if (currentScore > score) + { + score = currentScore; + currentHighestFeature = j; + } + }/*if j is unselected*/ + }/*for number of features*/ + + selectedFeatures[currentHighestFeature] = 1; + outputFeatures[i] = currentHighestFeature; + + }/*for the number of features to select*/ + + /*C++ indexes from 0 not 1, so we need to increment all the feature indices*/ + for (i = 0; i < k; i++) + { + outputFeatures[i] += 1; + }/*for number of selected features*/ + + FREE_FUNC(classMI); + FREE_FUNC(feature2D); + FREE_FUNC(featureMIMatrix); + FREE_FUNC(featureCMIMatrix); + FREE_FUNC(selectedFeatures); + + classMI = NULL; + feature2D = NULL; + featureMIMatrix = NULL; + featureCMIMatrix = NULL; + selectedFeatures = NULL; +}/*ICAP(int,int,int,double[][],double[],double[])*/ + diff --git a/FEAST/FEAST/FSToolbox/JMI.c b/FEAST/FEAST/FSToolbox/JMI.c new file mode 100644 index 0000000..f7574da --- /dev/null +++ b/FEAST/FEAST/FSToolbox/JMI.c @@ -0,0 +1,175 @@ +/******************************************************************************* +** JMI.c implements the JMI criterion from +** +** "Data Visualization and Feature Selection: New Algorithms for Nongaussian Data" +** H. Yang and J. Moody, NIPS (1999) +** +** Initial Version - 19/08/2010 +** Updated - 23/06/2011 +** +** Author - Adam Pocock +** +** Part of the Feature Selection Toolbox, please reference +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +** Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +** +** Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +** +** Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +** All rights reserved. +** +** Redistribution and use in source and binary forms, with or without modification, +** are permitted provided that the following conditions are met: +** +** - Redistributions of source code must retain the above copyright notice, this +** list of conditions and the following disclaimer. +** - Redistributions in binary form must reproduce the above copyright notice, +** this list of conditions and the following disclaimer in the documentation +** and/or other materials provided with the distribution. +** - Neither the name of The University of Manchester nor the names of its +** contributors may be used to endorse or promote products derived from this +** software without specific prior written permission. +** +** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +** ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +** WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +** DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +** ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +** (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +** LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +** ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +** SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +** +*******************************************************************************/ + +#include "FSAlgorithms.h" +#include "FSToolbox.h" + +/* MIToolbox includes */ +#include "MutualInformation.h" +#include "ArrayOperations.h" + +void JMI(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures) +{ + /*holds the class MI values*/ + double *classMI = (double *)CALLOC_FUNC(noOfFeatures,sizeof(double)); + + char *selectedFeatures = (char *)CALLOC_FUNC(noOfFeatures,sizeof(char)); + + /*holds the intra feature MI values*/ + int sizeOfMatrix = k*noOfFeatures; + double *featureMIMatrix = (double *)CALLOC_FUNC(sizeOfMatrix,sizeof(double)); + + double maxMI = 0.0; + int maxMICounter = -1; + + double **feature2D = (double**) CALLOC_FUNC(noOfFeatures,sizeof(double*)); + + double score, currentScore, totalFeatureMI; + int currentHighestFeature; + + double *mergedVector = (double *) CALLOC_FUNC(noOfSamples,sizeof(double)); + + int arrayPosition; + double mi, tripEntropy; + + int i,j,x; + + for(j = 0; j < noOfFeatures; j++) + { + feature2D[j] = featureMatrix + (int)j*noOfSamples; + } + + for (i = 0; i < sizeOfMatrix;i++) + { + featureMIMatrix[i] = -1; + }/*for featureMIMatrix - blank to -1*/ + + + for (i = 0; i < noOfFeatures;i++) + { + /*calculate mutual info + **double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength); + */ + classMI[i] = calculateMutualInformation(feature2D[i], classColumn, noOfSamples); + + if (classMI[i] > maxMI) + { + maxMI = classMI[i]; + maxMICounter = i; + }/*if bigger than current maximum*/ + }/*for noOfFeatures - filling classMI*/ + + selectedFeatures[maxMICounter] = 1; + outputFeatures[0] = maxMICounter; + + /***************************************************************************** + ** We have populated the classMI array, and selected the highest + ** MI feature as the first output feature + ** Now we move into the JMI algorithm + *****************************************************************************/ + + for (i = 1; i < k; i++) + { + score = 0.0; + currentHighestFeature = 0; + currentScore = 0.0; + totalFeatureMI = 0.0; + + for (j = 0; j < noOfFeatures; j++) + { + /*if we haven't selected j*/ + if (selectedFeatures[j] == 0) + { + currentScore = 0.0; + totalFeatureMI = 0.0; + + for (x = 0; x < i; x++) + { + arrayPosition = x*noOfFeatures + j; + if (featureMIMatrix[arrayPosition] == -1) + { + mergeArrays(feature2D[(int) outputFeatures[x]], feature2D[j],mergedVector,noOfSamples); + /*double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength);*/ + mi = calculateMutualInformation(mergedVector, classColumn, noOfSamples); + + featureMIMatrix[arrayPosition] = mi; + }/*if not already known*/ + currentScore += featureMIMatrix[arrayPosition]; + }/*for the number of already selected features*/ + + if (currentScore > score) + { + score = currentScore; + currentHighestFeature = j; + } + }/*if j is unselected*/ + }/*for number of features*/ + + selectedFeatures[currentHighestFeature] = 1; + outputFeatures[i] = currentHighestFeature; + + }/*for the number of features to select*/ + + for (i = 0; i < k; i++) + { + outputFeatures[i] += 1; /*C indexes from 0 not 1*/ + }/*for number of selected features*/ + + FREE_FUNC(classMI); + FREE_FUNC(feature2D); + FREE_FUNC(featureMIMatrix); + FREE_FUNC(mergedVector); + FREE_FUNC(selectedFeatures); + + classMI = NULL; + feature2D = NULL; + featureMIMatrix = NULL; + mergedVector = NULL; + selectedFeatures = NULL; + +}/*JMI(int,int,int,double[][],double[],double[])*/ + diff --git a/FEAST/FEAST/FSToolbox/MIM.m b/FEAST/FEAST/FSToolbox/MIM.m new file mode 100644 index 0000000..31695e4 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/MIM.m @@ -0,0 +1,17 @@ +function [selectedFeatures scoreVector] = MIM(k, data, labels)
+%function [selectedFeatures scoreVector] = MIM(k, data, labels)
+%
+%Mutual information Maximisation
+%
+% The license is in the license.txt provided.
+
+numf = size(data,2);
+classMI = zeros(numf,1);
+
+for n = 1 : numf
+ classMI(n) = mi(data(:,n),labels);
+end
+
+[scoreVector index] = sort(classMI,'descend');
+
+selectedFeatures = index(1:k);
diff --git a/FEAST/FEAST/FSToolbox/Makefile b/FEAST/FEAST/FSToolbox/Makefile new file mode 100644 index 0000000..c0806a9 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/Makefile @@ -0,0 +1,96 @@ +# makefile for FEAST +# Author: Adam Pocock, apocock@cs.man.ac.uk +# Created: 29/06/2011 +# +# Part of the Feature Selection Toolbox, please reference +# "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +# Information Feature Selection" +# G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +# Journal of Machine Learning Research (JMLR), 2012 +# +# Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +# +# Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +# All rights reserved. +# +# Redistribution and use in source and binary forms, with or without modification, +# are permitted provided that the following conditions are met: +# +# - Redistributions of source code must retain the above copyright notice, this +# list of conditions and the following disclaimer. +# - Redistributions in binary form must reproduce the above copyright notice, +# this list of conditions and the following disclaimer in the documentation +# and/or other materials provided with the distribution. +# - Neither the name of The University of Manchester nor the names of its +# contributors may be used to endorse or promote products derived from this +# software without specific prior written permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +# ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +# ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +CXXFLAGS = -O3 -fPIC +COMPILER = gcc +MITOOLBOXPATH = ../MIToolbox/ +objects = mRMR_D.o CMIM.o JMI.o DISR.o CondMI.o ICAP.o BetaGamma.o + +libFSToolbox.so : $(objects) + $(COMPILER) $(CXXFLAGS) -lMIToolbox -L$(MITOOLBOXPATH) -shared -o libFSToolbox.so $(objects) + +mRMR_D.o: mRMR_D.c + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c mRMR_D.c -I$(MITOOLBOXPATH) + +CMIM.o: CMIM.c + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c CMIM.c -I$(MITOOLBOXPATH) + +JMI.o: JMI.c + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c JMI.c -I$(MITOOLBOXPATH) + +DISR.o: DISR.c + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c DISR.c -I$(MITOOLBOXPATH) + +CondMI.o: CondMI.c + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c CondMI.c -I$(MITOOLBOXPATH) + +ICAP.o: ICAP.c + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c ICAP.c -I$(MITOOLBOXPATH) + +BetaGamma.o: BetaGamma.c + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c BetaGamma.c -I$(MITOOLBOXPATH) + +.PHONY : debug +debug: + $(MAKE) libFSToolbox.so "CXXFLAGS = -g -DDEBUG -fPIC" + +.PHONY : x86 +x86: + $(MAKE) libFSToolbox.so "CXXFLAGS = -O3 -fPIC -m32" + +.PHONY : x64 +x64: + $(MAKE) libFSToolbox.so "CXXFLAGS = -O3 -fPIC -m64" + +.PHONY : matlab +matlab: + mex -I$(MITOOLBOXPATH) FSToolboxMex.c BetaGamma.c CMIM.c CondMI.c DISR.c ICAP.c JMI.c mRMR_D.c $(MITOOLBOXPATH)MutualInformation.c $(MITOOLBOXPATH)Entropy.c $(MITOOLBOXPATH)CalculateProbability.c $(MITOOLBOXPATH)ArrayOperations.c + +.PHONY : matlab-debug +matlab-debug: + mex -g -I$(MITOOLBOXPATH) FSToolboxMex.c BetaGamma.c CMIM.c CondMI.c DISR.c ICAP.c JMI.c mRMR_D.c $(MITOOLBOXPATH)MutualInformation.c $(MITOOLBOXPATH)Entropy.c $(MITOOLBOXPATH)CalculateProbability.c $(MITOOLBOXPATH)ArrayOperations.c + +.PHONY : intel +intel: + $(MAKE) libFSToolbox.so "COMPILER = icc" "CXXFLAGS = -O2 -fPIC -xHost" + +.PHONY : clean +clean: + rm *.o + rm libFSToolbox.so + diff --git a/FEAST/FEAST/FSToolbox/README b/FEAST/FEAST/FSToolbox/README new file mode 100644 index 0000000..1aae2d7 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/README @@ -0,0 +1,80 @@ +FEAST v1.0 +A feature selection toolbox for C/C++ and MATLAB/OCTAVE + +FEAST provides implementations of common mutual information based filter +feature selection algorithms, and an implementation of RELIEF. All +functions expect discrete inputs (except RELIEF, which does not depend +on the MIToolbox), and they return the selected feature indices. These +implementations were developed to help our research into the similarities +between these algorithms, and our results are presented in the following paper: + + Conditional Likelihood Maximisation: A Unifying Framework for Mutual Information Feature Selection + G.Brown, A.Pocock, M.Lujan, M.-J.Zhao + Journal of Machine Learning Research (in press, to appear 2012) + +All FEAST code is licensed under the BSD 3-Clause License. +If you use these implementations for academic research please cite the paper above. + +Contains implementations of: + mim, mrmr, mifs, cmim, jmi, disr, cife, icap, condred, cmi, relief, fcbf, betagamma + +References for these algorithms are provided in the accompanying feast.bib file (in BibTeX format). + +MATLAB Example (using "data" as our feature matrix, and "labels" as the class label vector): + +>> size(data) +ans = + (569,30) %% denoting 569 examples, and 30 features + +>> selectedIndices = feast('jmi',5,data,labels) %% selecting the top 5 features using the jmi algorithm +selectedIndices = + + 28 + 21 + 8 + 27 + 23 + +>> selectedIndices = feast('mrmr',10,data,labels) %% selecting the top 10 features using the mrmr algorithm +selectedIndices = + + 28 + 24 + 22 + 8 + 27 + 21 + 29 + 4 + 7 + 25 + +>> selectedIndices = feast('mifs',5,data,labels,0.7) %% selecting the top 5 features using the mifs algorithm with beta = 0.7 +selectedIndices = + + 28 + 24 + 22 + 20 + 29 + +The library is written in ANSI C for compatibility with the MATLAB mex compiler, +except for MIM, FCBF and RELIEF, which are written in MATLAB/OCTAVE script. + +If you wish to use MIM in a C program you can use the BetaGamma function with +Beta = 0, Gamma = 0, as this is equivalent to MIM (but slower than the other implementation). +MIToolbox is required to compile these algorithms, and these implementations +supercede the example implementations given in that package (they have more robust behaviour +when used with unexpected inputs). + +MIToolbox can be found at: + http://www.cs.man.ac.uk/~gbrown/mitoolbox/ +and v1.03 is included in the ZIP for the FEAST package. + +Compilation instructions: + MATLAB/OCTAVE - run CompileFEAST.m, + Linux C shared library - use the included makefile + +Update History +08/11/2011 - v1.0 - Public Release to complement the JMLR publication. + diff --git a/FEAST/FEAST/FSToolbox/RELIEF.m b/FEAST/FEAST/FSToolbox/RELIEF.m new file mode 100644 index 0000000..194ce7b --- /dev/null +++ b/FEAST/FEAST/FSToolbox/RELIEF.m @@ -0,0 +1,61 @@ +% RELIEF - Kira & Rendell 1992 +% T is number of patterns to use +% Defaults to all patterns if not specified. +% +% The license is in the license.txt provided. +% +% function w = RELIEF( data, labels, T ) +% +function [w bestidx] = RELIEF ( data, labels, T ) + +if ~exist('T','var') + T=size(data,1); +end + +idx = randperm(length(labels)); +idx = idx(1:T); + +w = zeros(size(data,2),1); +for t = 1:T + + x = data(idx(t),:); + y = labels(idx(t)); + + %copy the x + protos = repmat(x, length(labels), 1); + %measure the distance from x to every other example + distances = [sqrt(sum((data-protos).^2,2)) labels]; + %sort them according to distances (find nearest neighbours) + [distances originalidx] = sortrows(distances,1); + + foundhit = false; hitidx=0; + foundmiss = false; missidx=0; + i=2; %start from the second one + while (~foundhit || ~foundmiss) + + if distances(i,2) == y + hitidx = originalidx(i); + foundhit = true; + end + if distances(i,2) ~= y + missidx = originalidx(i); + foundmiss = true; + end + + i=i+1; + + end + + alpha = 1/T; + for f = 1:size(data,2)%each feature + hitpenalty = (x(f)-data(hitidx,f)) / (max(data(:,f))-min(data(:,f))); + misspenalty = (x(f)-data(missidx,f)) / (max(data(:,f))-min(data(:,f))); + + w(f) = w(f) - alpha*hitpenalty^2 + alpha*misspenalty^2; + end + +end + +[~,bestidx] = sort(w,'descend'); + + diff --git a/FEAST/FEAST/FSToolbox/feast.bib b/FEAST/FEAST/FSToolbox/feast.bib new file mode 100644 index 0000000..2c58f0d --- /dev/null +++ b/FEAST/FEAST/FSToolbox/feast.bib @@ -0,0 +1,122 @@ +% MIM (Mutual Information Maximisation) +@inproceedings{MIM, + author = {David D. Lewis}, + title = {Feature Selection and Feature Extraction for Text Categorization}, + booktitle = {In Proceedings of Speech and Natural Language Workshop}, + year = {1992}, + pages = {212--217}, + publisher = {Morgan Kaufmann} +} + +% MIFS (Mutual Information Feature Selection ) +@article{MIFS, + author={Battiti, R.}, + journal={Neural Networks, IEEE Transactions on}, + title={Using mutual information for selecting features in supervised neural net learning}, + year={1994}, + month={jul}, + volume={5}, + number={4}, + pages={537 -550}, + ISSN={1045-9227} +} + +% mRMR (minimum Redundancy Maximum Relevance) +@article{mRMR, + title={Feature selection based on mutual information: criteria of max-dependency, max-relevance, and min-redundancy}, + author={Peng, H. and Long, F. and Ding, C.}, + journal={IEEE Transactions on pattern analysis and machine intelligence}, + pages={1226--1238}, + year={2005}, + publisher={Published by the IEEE Computer Society} +} + +% CMIM (Conditional Mutual Information Maximisation) +@article{CMIM, + author = {Fleuret, Fran\c{c}ois}, + title = {Fast Binary Feature Selection with Conditional Mutual Information}, + journal = {Journal of Machine Learning Research}, + volume = {5}, + month = {December}, + year = {2004}, + issn = {1532-4435}, + pages = {1531--1555}, + publisher = {JMLR.org} +} + +% JMI (Joint Mutual Information) +@inproceedings{JMI, + title={Feature selection based on joint mutual information}, + author={Yang, H. and Moody, J.}, + booktitle={Proceedings of International ICSC Symposium on Advances in Intelligent Data Analysis}, + pages={22--25}, + year={1999} +} + +% DISR (Double Input Symmetrical Relevance) +@incollection{DISR, + author = {Meyer, Patrick and Bontempi, Gianluca}, + title = {On the Use of Variable Complementarity for Feature Selection in Cancer Classification}, + booktitle = {Applications of Evolutionary Computing}, + publisher = {Springer Berlin / Heidelberg}, + pages = {91-102}, + volume = {3907}, + url = {http://dx.doi.org/10.1007/11732242_9}, + year = {2006} +} + +% ICAP (Interaction Capping) +@article{ICAP, + title={Machine learning based on attribute interactions}, + author={Jakulin, A.}, + journal={Fakulteta za racunalni{\v{s}}tvo in informatiko, Univerza v Ljubljani}, + year={2005} +} + +% CIFE (Conditional Informative Feature Extraction) +@incollection{CIFE + author = {Lin, Dahua and Tang, Xiaoou}, + title = {Conditional Infomax Learning: An Integrated Framework for Feature Extraction and Fusion}, + booktitle = {Computer Vision – ECCV 2006}, + series = {Lecture Notes in Computer Science}, + publisher = {Springer Berlin / Heidelberg}, + pages = {68-82}, + volume = {3951}, + url = {http://dx.doi.org/10.1007/11744023_6}, + year = {2006} +} + +% Beta Gamma Space +@inproceedings{BetaGamma, + title={A new perspective for information theoretic feature selection}, + author={Brown, G.}, + booktitle={12th International Conference on Artificial Intelligence and Statistics}, + volume={5}, + pages={49--56}, + year={2009} +} + +% FCBF (Fast Correlation-Based Filter) +@article{FCBF, + author = {Yu, Lei and Liu, Huan}, + title = {Efficient Feature Selection via Analysis of Relevance and Redundancy}, + journal = {Journal of Machine Learning Research}, + volume = {5}, + year = {2004}, + issn = {1532-4435}, + pages = {1205--1224}, + publisher = {JMLR.org} +} + +% RELIEF +@inproceedings{RELIEF, + author = {Kira, Kenji and Rendell, Larry A.}, + title = {The feature selection problem: traditional methods and a new algorithm}, + booktitle = {Proceedings of the tenth national conference on Artificial intelligence}, + series = {AAAI'92}, + year = {1992}, + pages = {129--134}, + publisher = {AAAI Press} +} + + diff --git a/FEAST/FEAST/FSToolbox/feast.m b/FEAST/FEAST/FSToolbox/feast.m new file mode 100644 index 0000000..96a685e --- /dev/null +++ b/FEAST/FEAST/FSToolbox/feast.m @@ -0,0 +1,100 @@ +function [selectedFeatures] = feast(criteria,numToSelect,data,labels,varargin) +%function [selectedFeatures] = feast(criteria,numToSelect,data,labels,varargin) +% +%Provides access to the feature selection algorithms in FSToolboxMex +% +%Expects the features to be columns of the data matrix, and +%requires that both the features and labels are integers. +% +%Algorithms are called as follows +% +%[selectedFeatures] = feast('algName',numToSelect,data,labels) +% where algName is: +% mim, mrmr, cmim, jmi, disr, cife, icap, condred, cmi, relief +% +%[selectedFeatures] = feast('algName',numToSelect,data,labels,beta) +% where algName is: +% mifs (defaults to beta = 1.0 if unspecified) +% +%[selectedFeatures] = feast('algName',numToSelect,data,labels,beta,gamma) +% where algName is: +% betagamma +%[selectedFeatures] = feast('algName',numToSelect,data,labels,threshold) +% where algName is: +% fcbf (note this ignores the numToSelect) +% +% The license is in the license.txt provided. + + +%Internal FSToolbox Criteria to number mapping +%MIFS = 1 +%mRMR = 2 +%CMIM = 3 +%JMI = 4 +%DISR = 5 +%CIFE = 6 +%ICAP = 7 +%CondRed = 8 +%BetaGamma = 9 +%CMI = 10 +% + +if ((numToSelect < 1) || (numToSelect > size(data,2))) + error(['You have requested ' num2str(numToSelect) ' features, which is not possible']); +end + +finiteDataCount = sum(sum(isfinite(data))); +finiteLabelsCount = sum(sum(isfinite(labels))); + +totalData = numel(data); +totalLabels = numel(labels); + +if ((finiteDataCount ~= totalData) || (finiteLabelsCount ~= totalLabels)) + error(['Some elements are NaNs or infinite. Please check your data']); +end + +if (strcmpi(criteria,'mim')) + selectedFeatures = MIM(numToSelect,data,labels); +elseif (strcmpi(criteria,'mifs')) + if (nargin == 4) + beta = 1; + else + beta = varargin{1}; + end + selectedFeatures = FSToolboxMex(1,numToSelect,data,labels,beta); +elseif (strcmpi(criteria,'mrmr')) + selectedFeatures = FSToolboxMex(2,numToSelect,data,labels); +elseif (strcmpi(criteria,'cmim')) + selectedFeatures = FSToolboxMex(3,numToSelect,data,labels); +elseif (strcmpi(criteria,'jmi')) + selectedFeatures = FSToolboxMex(4,numToSelect,data,labels); +elseif (strcmpi(criteria,'disr')) + selectedFeatures = FSToolboxMex(5,numToSelect,data,labels); +elseif ((strcmpi(criteria,'cife')) || (strcmpi(criteria,'fou'))) + selectedFeatures = FSToolboxMex(6,numToSelect,data,labels); +elseif (strcmpi(criteria,'icap')) + selectedFeatures = FSToolboxMex(7,numToSelect,data,labels); +elseif (strcmpi(criteria,'condred')) + selectedFeatures = FSToolboxMex(8,numToSelect,data,labels); +elseif (strcmpi(criteria,'betagamma')) + if (nargin ~= 6) + error('BetaGamma criteria expects a beta and a gamma'); + else + beta = varargin{1}; + gamma = varargin{2}; + end + selectedFeatures = FSToolboxMex(9,numToSelect,data,labels,beta,gamma); +elseif (strcmpi(criteria,'cmi')) + selectedFeatures = FSToolboxMex(10,numToSelect,data,labels); +elseif (strcmpi(criteria,'fcbf')) + if (nargin == 4) + error('Threshold for FCBF not supplied'); + else + selectedFeatures = FCBF(data,labels,varargin{1}); + end +elseif (strcmpi(criteria,'relief')) + [tmp selectedFeatures] = RELIEF(data,labels); +else + selectedFeatures = []; + disp(['Unrecognised criteria ' criteria]); +end diff --git a/FEAST/FEAST/FSToolbox/license.txt b/FEAST/FEAST/FSToolbox/license.txt new file mode 100644 index 0000000..798960e --- /dev/null +++ b/FEAST/FEAST/FSToolbox/license.txt @@ -0,0 +1,32 @@ +This is FEAST (a FEAture Selection Toolbox for C and MATLAB) +if you use this code in academic work could you please reference: +"Conditional Likelihood Maximisation: A Unifying Framework for Mutual +Information Feature Selection" +G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +Journal of Machine Learning Research (JMLR), 2011 + +Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +All rights reserved. + +Redistribution and use in source and binary forms, with or without modification, +are permitted provided that the following conditions are met: + + - Redistributions of source code must retain the above copyright notice, this + list of conditions and the following disclaimer. + - Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the documentation + and/or other materials provided with the distribution. + - Neither the name of The University of Manchester nor the names of its + contributors may be used to endorse or promote products derived from this + software without specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/FEAST/FEAST/FSToolbox/mRMR_D.c b/FEAST/FEAST/FSToolbox/mRMR_D.c new file mode 100644 index 0000000..6710866 --- /dev/null +++ b/FEAST/FEAST/FSToolbox/mRMR_D.c @@ -0,0 +1,169 @@ +/******************************************************************************* +** mRMR_D.c implements the minimum Relevance Maximum Redundancy criterion +** using the difference variant, from +** +** "Feature Selection Based on Mutual Information: Criteria of Max-Dependency, Max-Relevance, and Min-Redundancy" +** H. Peng et al. IEEE PAMI (2005) +** +** Initial Version - 13/06/2008 +** Updated - 23/06/2011 +** +** Author - Adam Pocock +** +** Part of the Feature Selection Toolbox, please reference +** "Conditional Likelihood Maximisation: A Unifying Framework for Mutual +** Information Feature Selection" +** G. Brown, A. Pocock, M.-J. Zhao, M. Lujan +** Journal of Machine Learning Research (JMLR), 2011 +** +** Please check www.cs.manchester.ac.uk/~gbrown/fstoolbox for updates. +** +** Copyright (c) 2010-2011, A. Pocock, G. Brown, The University of Manchester +** All rights reserved. +** +** Redistribution and use in source and binary forms, with or without modification, +** are permitted provided that the following conditions are met: +** +** - Redistributions of source code must retain the above copyright notice, this +** list of conditions and the following disclaimer. +** - Redistributions in binary form must reproduce the above copyright notice, +** this list of conditions and the following disclaimer in the documentation +** and/or other materials provided with the distribution. +** - Neither the name of The University of Manchester nor the names of its +** contributors may be used to endorse or promote products derived from this +** software without specific prior written permission. +** +** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +** ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +** WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +** DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR +** ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +** (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +** LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON +** ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +** (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +** SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +** +*******************************************************************************/ + +#include "FSAlgorithms.h" +#include "FSToolbox.h" + +/* MIToolbox includes */ +#include "MutualInformation.h" + +void mRMR_D(int k, int noOfSamples, int noOfFeatures, double *featureMatrix, double *classColumn, double *outputFeatures) +{ + double **feature2D = (double**) CALLOC_FUNC(noOfFeatures,sizeof(double*)); + /*holds the class MI values*/ + double *classMI = (double *)CALLOC_FUNC(noOfFeatures,sizeof(double)); + int *selectedFeatures = (int *)CALLOC_FUNC(noOfFeatures,sizeof(int)); + /*holds the intra feature MI values*/ + int sizeOfMatrix = k*noOfFeatures; + double *featureMIMatrix = (double *)CALLOC_FUNC(sizeOfMatrix,sizeof(double)); + + double maxMI = 0.0; + int maxMICounter = -1; + + /*init variables*/ + + double score, currentScore, totalFeatureMI; + int currentHighestFeature; + + int arrayPosition, i, j, x; + + for(j = 0; j < noOfFeatures; j++) + { + feature2D[j] = featureMatrix + (int)j*noOfSamples; + } + + for (i = 0; i < sizeOfMatrix;i++) + { + featureMIMatrix[i] = -1; + }/*for featureMIMatrix - blank to -1*/ + + + for (i = 0; i < noOfFeatures;i++) + { + classMI[i] = calculateMutualInformation(feature2D[i], classColumn, noOfSamples); + if (classMI[i] > maxMI) + { + maxMI = classMI[i]; + maxMICounter = i; + }/*if bigger than current maximum*/ + }/*for noOfFeatures - filling classMI*/ + + selectedFeatures[maxMICounter] = 1; + outputFeatures[0] = maxMICounter; + + /************* + ** Now we have populated the classMI array, and selected the highest + ** MI feature as the first output feature + ** Now we move into the mRMR-D algorithm + *************/ + + for (i = 1; i < k; i++) + { + /**************************************************** + ** to ensure it selects some features + **if this is zero then it will not pick features where the redundancy is greater than the + **relevance + ****************************************************/ + score = -HUGE_VAL; + currentHighestFeature = 0; + currentScore = 0.0; + totalFeatureMI = 0.0; + + for (j = 0; j < noOfFeatures; j++) + { + /*if we haven't selected j*/ + if (selectedFeatures[j] == 0) + { + currentScore = classMI[j]; + totalFeatureMI = 0.0; + + for (x = 0; x < i; x++) + { + arrayPosition = x*noOfFeatures + j; + if (featureMIMatrix[arrayPosition] == -1) + { + /*work out intra MI*/ + + /*double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength);*/ + featureMIMatrix[arrayPosition] = calculateMutualInformation(feature2D[(int) outputFeatures[x]], feature2D[j], noOfSamples); + } + + totalFeatureMI += featureMIMatrix[arrayPosition]; + }/*for the number of already selected features*/ + + currentScore -= (totalFeatureMI/i); + if (currentScore > score) + { + score = currentScore; + currentHighestFeature = j; + } + }/*if j is unselected*/ + }/*for number of features*/ + + selectedFeatures[currentHighestFeature] = 1; + outputFeatures[i] = currentHighestFeature; + + }/*for the number of features to select*/ + + for (i = 0; i < k; i++) + { + outputFeatures[i] += 1; /*C indexes from 0 not 1*/ + }/*for number of selected features*/ + + FREE_FUNC(classMI); + FREE_FUNC(feature2D); + FREE_FUNC(featureMIMatrix); + FREE_FUNC(selectedFeatures); + + classMI = NULL; + feature2D = NULL; + featureMIMatrix = NULL; + selectedFeatures = NULL; + +}/*mRMR(int,int,int,double[][],double[],double[])*/ + diff --git a/FEAST/FEAST/MIToolbox/ArrayOperations.c b/FEAST/FEAST/MIToolbox/ArrayOperations.c new file mode 100644 index 0000000..00a8324 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/ArrayOperations.c @@ -0,0 +1,288 @@ +/******************************************************************************* +** ArrayOperations.cpp +** Part of the mutual information toolbox +** +** Contains functions to floor arrays, and to merge arrays into a joint +** state. +** +** Author: Adam Pocock +** Created 17/2/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#include "MIToolbox.h" +#include "ArrayOperations.h" + +void printDoubleVector(double *vector, int vectorLength) +{ + int i; + for (i = 0; i < vectorLength; i++) + { + if (vector[i] > 0) + printf("Val at i=%d, is %f\n",i,vector[i]); + }/*for number of items in vector*/ +}/*printDoubleVector(double*,int)*/ + +void printIntVector(int *vector, int vectorLength) +{ + int i; + for (i = 0; i < vectorLength; i++) + { + printf("Val at i=%d, is %d\n",i,vector[i]); + }/*for number of items in vector*/ +}/*printIntVector(int*,int)*/ + +int numberOfUniqueValues(double *featureVector, int vectorLength) +{ + int uniqueValues = 0; + double *valuesArray = (double *) CALLOC_FUNC(vectorLength,sizeof(double)); + + int found = 0; + int j = 0; + int i; + + for (i = 0; i < vectorLength; i++) + { + found = 0; + j = 0; + while ((j < uniqueValues) && (found == 0)) + { + if (valuesArray[j] == featureVector[i]) + { + found = 1; + featureVector[i] = (double) (j+1); + } + j++; + } + if (!found) + { + valuesArray[uniqueValues] = featureVector[i]; + uniqueValues++; + featureVector[i] = (double) uniqueValues; + } + }/*for vectorlength*/ + + FREE_FUNC(valuesArray); + valuesArray = NULL; + + return uniqueValues; +}/*numberOfUniqueValues(double*,int)*/ + +/******************************************************************************* +** normaliseArray takes an input vector and writes an output vector +** which is a normalised version of the input, and returns the number of states +** A normalised array has min value = 0, max value = number of states +** and all values are integers +** +** length(inputVector) == length(outputVector) == vectorLength otherwise there +** is a memory leak +*******************************************************************************/ +int normaliseArray(double *inputVector, int *outputVector, int vectorLength) +{ + int minVal = 0; + int maxVal = 0; + int currentValue; + int i; + + if (vectorLength > 0) + { + minVal = (int) floor(inputVector[0]); + maxVal = (int) floor(inputVector[0]); + + for (i = 0; i < vectorLength; i++) + { + currentValue = (int) floor(inputVector[i]); + outputVector[i] = currentValue; + + if (currentValue < minVal) + { + minVal = currentValue; + } + + if (currentValue > maxVal) + { + maxVal = currentValue; + } + }/*for loop over vector*/ + + for (i = 0; i < vectorLength; i++) + { + outputVector[i] = outputVector[i] - minVal; + } + + maxVal = (maxVal - minVal) + 1; + } + + return maxVal; +}/*normaliseArray(double*,double*,int)*/ + + +/******************************************************************************* +** mergeArrays takes in two arrays and writes the joint state of those arrays +** to the output vector, returning the number of joint states +** +** the length of the vectors must be the same and equal to vectorLength +** outputVector must be malloc'd before calling this function +*******************************************************************************/ +int mergeArrays(double *firstVector, double *secondVector, double *outputVector, int vectorLength) +{ + int *firstNormalisedVector; + int *secondNormalisedVector; + int firstNumStates; + int secondNumStates; + int i; + int *stateMap; + int stateCount; + int curIndex; + + firstNormalisedVector = (int *) CALLOC_FUNC(vectorLength,sizeof(int)); + secondNormalisedVector = (int *) CALLOC_FUNC(vectorLength,sizeof(int)); + + firstNumStates = normaliseArray(firstVector,firstNormalisedVector,vectorLength); + secondNumStates = normaliseArray(secondVector,secondNormalisedVector,vectorLength); + + /* + ** printVector(firstNormalisedVector,vectorLength); + ** printVector(secondNormalisedVector,vectorLength); + */ + stateMap = (int *) CALLOC_FUNC(firstNumStates*secondNumStates,sizeof(int)); + stateCount = 1; + for (i = 0; i < vectorLength; i++) + { + curIndex = firstNormalisedVector[i] + (secondNormalisedVector[i] * firstNumStates); + if (stateMap[curIndex] == 0) + { + stateMap[curIndex] = stateCount; + stateCount++; + } + outputVector[i] = stateMap[curIndex]; + } + + FREE_FUNC(firstNormalisedVector); + FREE_FUNC(secondNormalisedVector); + FREE_FUNC(stateMap); + + firstNormalisedVector = NULL; + secondNormalisedVector = NULL; + stateMap = NULL; + + /*printVector(outputVector,vectorLength);*/ + return stateCount; +}/*mergeArrays(double *,double *,double *, int, bool)*/ + +int mergeArraysArities(double *firstVector, int numFirstStates, double *secondVector, int numSecondStates, double *outputVector, int vectorLength) +{ + int *firstNormalisedVector; + int *secondNormalisedVector; + int i; + int totalStates; + int firstStateCheck, secondStateCheck; + + firstNormalisedVector = (int *) CALLOC_FUNC(vectorLength,sizeof(int)); + secondNormalisedVector = (int *) CALLOC_FUNC(vectorLength,sizeof(int)); + + firstStateCheck = normaliseArray(firstVector,firstNormalisedVector,vectorLength); + secondStateCheck = normaliseArray(secondVector,secondNormalisedVector,vectorLength); + + if ((firstStateCheck <= numFirstStates) && (secondStateCheck <= numSecondStates)) + { + for (i = 0; i < vectorLength; i++) + { + outputVector[i] = firstNormalisedVector[i] + (secondNormalisedVector[i] * numFirstStates) + 1; + } + totalStates = numFirstStates * numSecondStates; + } + else + { + totalStates = -1; + } + + FREE_FUNC(firstNormalisedVector); + FREE_FUNC(secondNormalisedVector); + + firstNormalisedVector = NULL; + secondNormalisedVector = NULL; + + return totalStates; +}/*mergeArraysArities(double *,int,double *,int,double *,int)*/ + +int mergeMultipleArrays(double *inputMatrix, double *outputVector, int matrixWidth, int vectorLength) +{ + int i = 0; + int currentIndex; + int currentNumStates; + int *normalisedVector; + + if (matrixWidth > 1) + { + currentNumStates = mergeArrays(inputMatrix, (inputMatrix + vectorLength), outputVector,vectorLength); + for (i = 2; i < matrixWidth; i++) + { + currentIndex = i * vectorLength; + currentNumStates = mergeArrays(outputVector,(inputMatrix + currentIndex),outputVector,vectorLength); + } + } + else + { + normalisedVector = (int *) CALLOC_FUNC(vectorLength,sizeof(int)); + currentNumStates = normaliseArray(inputMatrix,normalisedVector,vectorLength); + for (i = 0; i < vectorLength; i++) + { + outputVector[i] = inputMatrix[i]; + } + } + + return currentNumStates; +}/*mergeMultipleArrays(double *, double *, int, int, bool)*/ + + +int mergeMultipleArraysArities(double *inputMatrix, double *outputVector, int matrixWidth, int *arities, int vectorLength) +{ + int i = 0; + int currentIndex; + int currentNumStates; + int *normalisedVector; + + if (matrixWidth > 1) + { + currentNumStates = mergeArraysArities(inputMatrix, arities[0], (inputMatrix + vectorLength), arities[1], outputVector,vectorLength); + for (i = 2; i < matrixWidth; i++) + { + currentIndex = i * vectorLength; + currentNumStates = mergeArraysArities(outputVector,currentNumStates,(inputMatrix + currentIndex),arities[i],outputVector,vectorLength); + if (currentNumStates == -1) + break; + } + } + else + { + normalisedVector = (int *) CALLOC_FUNC(vectorLength,sizeof(int)); + currentNumStates = normaliseArray(inputMatrix,normalisedVector,vectorLength); + for (i = 0; i < vectorLength; i++) + { + outputVector[i] = inputMatrix[i]; + } + } + + return currentNumStates; +}/*mergeMultipleArraysArities(double *, double *, int, int, bool)*/ + + diff --git a/FEAST/FEAST/MIToolbox/ArrayOperations.h b/FEAST/FEAST/MIToolbox/ArrayOperations.h new file mode 100644 index 0000000..3cc9025 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/ArrayOperations.h @@ -0,0 +1,88 @@ +/******************************************************************************* +** ArrayOperations.h +** Part of the mutual information toolbox +** +** Contains functions to floor arrays, and to merge arrays into a joint +** state. +** +** Author: Adam Pocock +** Created 17/2/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#ifndef __ArrayOperations_H +#define __ArrayOperations_H + +#ifdef __cplusplus +extern "C" { +#endif + +/******************************************************************************* +** Simple print function for debugging +*******************************************************************************/ +void printDoubleVector(double *vector, int vectorLength); + +void printIntVector(int *vector, int vectorLength); + +/******************************************************************************* +** numberOfUniqueValues finds the number of unique values in an array by +** repeatedly iterating through the array and checking if a value has been +** seen previously +*******************************************************************************/ +int numberOfUniqueValues(double *featureVector, int vectorLength); + +/******************************************************************************* +** normaliseArray takes an input vector and writes an output vector +** which is a normalised version of the input, and returns the number of states +** A normalised array has min value = 0, max value = number of states +** and all values are integers +** +** length(inputVector) == length(outputVector) == vectorLength otherwise there +** is a memory leak +*******************************************************************************/ +int normaliseArray(double *inputVector, int *outputVector, int vectorLength); + +/******************************************************************************* +** mergeArrays takes in two arrays and writes the joint state of those arrays +** to the output vector +** +** the length of the vectors must be the same and equal to vectorLength +*******************************************************************************/ +int mergeArrays(double *firstVector, double *secondVector, double *outputVector, int vectorLength); +int mergeArraysArities(double *firstVector, int numFirstStates, double *secondVector, int numSecondStates, double *outputVector, int vectorLength); + +/******************************************************************************* +** mergeMultipleArrays takes in a matrix and repeatedly merges the matrix using +** merge arrays and writes the joint state of that matrix +** to the output vector +** +** the length of the vectors must be the same and equal to vectorLength +** matrixWidth = the number of columns in the matrix +*******************************************************************************/ +int mergeMultipleArrays(double *inputMatrix, double *outputVector, int matrixWidth, int vectorLength); +int mergeMultipleArraysArities(double *inputMatrix, double *outputVector, int matrixWidth, int *arities, int vectorLength); + +#ifdef __cplusplus +} +#endif + +#endif + diff --git a/FEAST/FEAST/MIToolbox/COPYING b/FEAST/FEAST/MIToolbox/COPYING new file mode 100644 index 0000000..94a9ed0 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/COPYING @@ -0,0 +1,674 @@ + GNU GENERAL PUBLIC LICENSE + Version 3, 29 June 2007 + + Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/> + Everyone is permitted to copy and distribute verbatim copies + of this license document, but changing it is not allowed. + + Preamble + + The GNU General Public License is a free, copyleft license for +software and other kinds of works. + + The licenses for most software and other practical works are designed +to take away your freedom to share and change the works. By contrast, +the GNU General Public License is intended to guarantee your freedom to +share and change all versions of a program--to make sure it remains free +software for all its users. We, the Free Software Foundation, use the +GNU General Public License for most of our software; it applies also to +any other work released this way by its authors. You can apply it to +your programs, too. + + When we speak of free software, we are referring to freedom, not +price. Our General Public Licenses are designed to make sure that you +have the freedom to distribute copies of free software (and charge for +them if you wish), that you receive source code or can get it if you +want it, that you can change the software or use pieces of it in new +free programs, and that you know you can do these things. + + To protect your rights, we need to prevent others from denying you +these rights or asking you to surrender the rights. Therefore, you have +certain responsibilities if you distribute copies of the software, or if +you modify it: responsibilities to respect the freedom of others. + + For example, if you distribute copies of such a program, whether +gratis or for a fee, you must pass on to the recipients the same +freedoms that you received. You must make sure that they, too, receive +or can get the source code. And you must show them these terms so they +know their rights. + + Developers that use the GNU GPL protect your rights with two steps: +(1) assert copyright on the software, and (2) offer you this License +giving you legal permission to copy, distribute and/or modify it. + + For the developers' and authors' protection, the GPL clearly explains +that there is no warranty for this free software. For both users' and +authors' sake, the GPL requires that modified versions be marked as +changed, so that their problems will not be attributed erroneously to +authors of previous versions. + + Some devices are designed to deny users access to install or run +modified versions of the software inside them, although the manufacturer +can do so. This is fundamentally incompatible with the aim of +protecting users' freedom to change the software. The systematic +pattern of such abuse occurs in the area of products for individuals to +use, which is precisely where it is most unacceptable. Therefore, we +have designed this version of the GPL to prohibit the practice for those +products. If such problems arise substantially in other domains, we +stand ready to extend this provision to those domains in future versions +of the GPL, as needed to protect the freedom of users. + + Finally, every program is threatened constantly by software patents. +States should not allow patents to restrict development and use of +software on general-purpose computers, but in those that do, we wish to +avoid the special danger that patents applied to a free program could +make it effectively proprietary. To prevent this, the GPL assures that +patents cannot be used to render the program non-free. + + The precise terms and conditions for copying, distribution and +modification follow. + + TERMS AND CONDITIONS + + 0. Definitions. + + "This License" refers to version 3 of the GNU General Public License. + + "Copyright" also means copyright-like laws that apply to other kinds of +works, such as semiconductor masks. + + "The Program" refers to any copyrightable work licensed under this +License. Each licensee is addressed as "you". "Licensees" and +"recipients" may be individuals or organizations. + + To "modify" a work means to copy from or adapt all or part of the work +in a fashion requiring copyright permission, other than the making of an +exact copy. The resulting work is called a "modified version" of the +earlier work or a work "based on" the earlier work. + + A "covered work" means either the unmodified Program or a work based +on the Program. + + To "propagate" a work means to do anything with it that, without +permission, would make you directly or secondarily liable for +infringement under applicable copyright law, except executing it on a +computer or modifying a private copy. Propagation includes copying, +distribution (with or without modification), making available to the +public, and in some countries other activities as well. + + To "convey" a work means any kind of propagation that enables other +parties to make or receive copies. Mere interaction with a user through +a computer network, with no transfer of a copy, is not conveying. + + An interactive user interface displays "Appropriate Legal Notices" +to the extent that it includes a convenient and prominently visible +feature that (1) displays an appropriate copyright notice, and (2) +tells the user that there is no warranty for the work (except to the +extent that warranties are provided), that licensees may convey the +work under this License, and how to view a copy of this License. If +the interface presents a list of user commands or options, such as a +menu, a prominent item in the list meets this criterion. + + 1. Source Code. + + The "source code" for a work means the preferred form of the work +for making modifications to it. "Object code" means any non-source +form of a work. + + A "Standard Interface" means an interface that either is an official +standard defined by a recognized standards body, or, in the case of +interfaces specified for a particular programming language, one that +is widely used among developers working in that language. + + The "System Libraries" of an executable work include anything, other +than the work as a whole, that (a) is included in the normal form of +packaging a Major Component, but which is not part of that Major +Component, and (b) serves only to enable use of the work with that +Major Component, or to implement a Standard Interface for which an +implementation is available to the public in source code form. A +"Major Component", in this context, means a major essential component +(kernel, window system, and so on) of the specific operating system +(if any) on which the executable work runs, or a compiler used to +produce the work, or an object code interpreter used to run it. + + The "Corresponding Source" for a work in object code form means all +the source code needed to generate, install, and (for an executable +work) run the object code and to modify the work, including scripts to +control those activities. However, it does not include the work's +System Libraries, or general-purpose tools or generally available free +programs which are used unmodified in performing those activities but +which are not part of the work. For example, Corresponding Source +includes interface definition files associated with source files for +the work, and the source code for shared libraries and dynamically +linked subprograms that the work is specifically designed to require, +such as by intimate data communication or control flow between those +subprograms and other parts of the work. + + The Corresponding Source need not include anything that users +can regenerate automatically from other parts of the Corresponding +Source. + + The Corresponding Source for a work in source code form is that +same work. + + 2. Basic Permissions. + + All rights granted under this License are granted for the term of +copyright on the Program, and are irrevocable provided the stated +conditions are met. This License explicitly affirms your unlimited +permission to run the unmodified Program. The output from running a +covered work is covered by this License only if the output, given its +content, constitutes a covered work. This License acknowledges your +rights of fair use or other equivalent, as provided by copyright law. + + You may make, run and propagate covered works that you do not +convey, without conditions so long as your license otherwise remains +in force. You may convey covered works to others for the sole purpose +of having them make modifications exclusively for you, or provide you +with facilities for running those works, provided that you comply with +the terms of this License in conveying all material for which you do +not control copyright. Those thus making or running the covered works +for you must do so exclusively on your behalf, under your direction +and control, on terms that prohibit them from making any copies of +your copyrighted material outside their relationship with you. + + Conveying under any other circumstances is permitted solely under +the conditions stated below. Sublicensing is not allowed; section 10 +makes it unnecessary. + + 3. Protecting Users' Legal Rights From Anti-Circumvention Law. + + No covered work shall be deemed part of an effective technological +measure under any applicable law fulfilling obligations under article +11 of the WIPO copyright treaty adopted on 20 December 1996, or +similar laws prohibiting or restricting circumvention of such +measures. + + When you convey a covered work, you waive any legal power to forbid +circumvention of technological measures to the extent such circumvention +is effected by exercising rights under this License with respect to +the covered work, and you disclaim any intention to limit operation or +modification of the work as a means of enforcing, against the work's +users, your or third parties' legal rights to forbid circumvention of +technological measures. + + 4. Conveying Verbatim Copies. + + You may convey verbatim copies of the Program's source code as you +receive it, in any medium, provided that you conspicuously and +appropriately publish on each copy an appropriate copyright notice; +keep intact all notices stating that this License and any +non-permissive terms added in accord with section 7 apply to the code; +keep intact all notices of the absence of any warranty; and give all +recipients a copy of this License along with the Program. + + You may charge any price or no price for each copy that you convey, +and you may offer support or warranty protection for a fee. + + 5. Conveying Modified Source Versions. + + You may convey a work based on the Program, or the modifications to +produce it from the Program, in the form of source code under the +terms of section 4, provided that you also meet all of these conditions: + + a) The work must carry prominent notices stating that you modified + it, and giving a relevant date. + + b) The work must carry prominent notices stating that it is + released under this License and any conditions added under section + 7. This requirement modifies the requirement in section 4 to + "keep intact all notices". + + c) You must license the entire work, as a whole, under this + License to anyone who comes into possession of a copy. This + License will therefore apply, along with any applicable section 7 + additional terms, to the whole of the work, and all its parts, + regardless of how they are packaged. This License gives no + permission to license the work in any other way, but it does not + invalidate such permission if you have separately received it. + + d) If the work has interactive user interfaces, each must display + Appropriate Legal Notices; however, if the Program has interactive + interfaces that do not display Appropriate Legal Notices, your + work need not make them do so. + + A compilation of a covered work with other separate and independent +works, which are not by their nature extensions of the covered work, +and which are not combined with it such as to form a larger program, +in or on a volume of a storage or distribution medium, is called an +"aggregate" if the compilation and its resulting copyright are not +used to limit the access or legal rights of the compilation's users +beyond what the individual works permit. Inclusion of a covered work +in an aggregate does not cause this License to apply to the other +parts of the aggregate. + + 6. Conveying Non-Source Forms. + + You may convey a covered work in object code form under the terms +of sections 4 and 5, provided that you also convey the +machine-readable Corresponding Source under the terms of this License, +in one of these ways: + + a) Convey the object code in, or embodied in, a physical product + (including a physical distribution medium), accompanied by the + Corresponding Source fixed on a durable physical medium + customarily used for software interchange. + + b) Convey the object code in, or embodied in, a physical product + (including a physical distribution medium), accompanied by a + written offer, valid for at least three years and valid for as + long as you offer spare parts or customer support for that product + model, to give anyone who possesses the object code either (1) a + copy of the Corresponding Source for all the software in the + product that is covered by this License, on a durable physical + medium customarily used for software interchange, for a price no + more than your reasonable cost of physically performing this + conveying of source, or (2) access to copy the + Corresponding Source from a network server at no charge. + + c) Convey individual copies of the object code with a copy of the + written offer to provide the Corresponding Source. This + alternative is allowed only occasionally and noncommercially, and + only if you received the object code with such an offer, in accord + with subsection 6b. + + d) Convey the object code by offering access from a designated + place (gratis or for a charge), and offer equivalent access to the + Corresponding Source in the same way through the same place at no + further charge. You need not require recipients to copy the + Corresponding Source along with the object code. If the place to + copy the object code is a network server, the Corresponding Source + may be on a different server (operated by you or a third party) + that supports equivalent copying facilities, provided you maintain + clear directions next to the object code saying where to find the + Corresponding Source. Regardless of what server hosts the + Corresponding Source, you remain obligated to ensure that it is + available for as long as needed to satisfy these requirements. + + e) Convey the object code using peer-to-peer transmission, provided + you inform other peers where the object code and Corresponding + Source of the work are being offered to the general public at no + charge under subsection 6d. + + A separable portion of the object code, whose source code is excluded +from the Corresponding Source as a System Library, need not be +included in conveying the object code work. + + A "User Product" is either (1) a "consumer product", which means any +tangible personal property which is normally used for personal, family, +or household purposes, or (2) anything designed or sold for incorporation +into a dwelling. In determining whether a product is a consumer product, +doubtful cases shall be resolved in favor of coverage. For a particular +product received by a particular user, "normally used" refers to a +typical or common use of that class of product, regardless of the status +of the particular user or of the way in which the particular user +actually uses, or expects or is expected to use, the product. A product +is a consumer product regardless of whether the product has substantial +commercial, industrial or non-consumer uses, unless such uses represent +the only significant mode of use of the product. + + "Installation Information" for a User Product means any methods, +procedures, authorization keys, or other information required to install +and execute modified versions of a covered work in that User Product from +a modified version of its Corresponding Source. The information must +suffice to ensure that the continued functioning of the modified object +code is in no case prevented or interfered with solely because +modification has been made. + + If you convey an object code work under this section in, or with, or +specifically for use in, a User Product, and the conveying occurs as +part of a transaction in which the right of possession and use of the +User Product is transferred to the recipient in perpetuity or for a +fixed term (regardless of how the transaction is characterized), the +Corresponding Source conveyed under this section must be accompanied +by the Installation Information. But this requirement does not apply +if neither you nor any third party retains the ability to install +modified object code on the User Product (for example, the work has +been installed in ROM). + + The requirement to provide Installation Information does not include a +requirement to continue to provide support service, warranty, or updates +for a work that has been modified or installed by the recipient, or for +the User Product in which it has been modified or installed. Access to a +network may be denied when the modification itself materially and +adversely affects the operation of the network or violates the rules and +protocols for communication across the network. + + Corresponding Source conveyed, and Installation Information provided, +in accord with this section must be in a format that is publicly +documented (and with an implementation available to the public in +source code form), and must require no special password or key for +unpacking, reading or copying. + + 7. Additional Terms. + + "Additional permissions" are terms that supplement the terms of this +License by making exceptions from one or more of its conditions. +Additional permissions that are applicable to the entire Program shall +be treated as though they were included in this License, to the extent +that they are valid under applicable law. If additional permissions +apply only to part of the Program, that part may be used separately +under those permissions, but the entire Program remains governed by +this License without regard to the additional permissions. + + When you convey a copy of a covered work, you may at your option +remove any additional permissions from that copy, or from any part of +it. (Additional permissions may be written to require their own +removal in certain cases when you modify the work.) You may place +additional permissions on material, added by you to a covered work, +for which you have or can give appropriate copyright permission. + + Notwithstanding any other provision of this License, for material you +add to a covered work, you may (if authorized by the copyright holders of +that material) supplement the terms of this License with terms: + + a) Disclaiming warranty or limiting liability differently from the + terms of sections 15 and 16 of this License; or + + b) Requiring preservation of specified reasonable legal notices or + author attributions in that material or in the Appropriate Legal + Notices displayed by works containing it; or + + c) Prohibiting misrepresentation of the origin of that material, or + requiring that modified versions of such material be marked in + reasonable ways as different from the original version; or + + d) Limiting the use for publicity purposes of names of licensors or + authors of the material; or + + e) Declining to grant rights under trademark law for use of some + trade names, trademarks, or service marks; or + + f) Requiring indemnification of licensors and authors of that + material by anyone who conveys the material (or modified versions of + it) with contractual assumptions of liability to the recipient, for + any liability that these contractual assumptions directly impose on + those licensors and authors. + + All other non-permissive additional terms are considered "further +restrictions" within the meaning of section 10. If the Program as you +received it, or any part of it, contains a notice stating that it is +governed by this License along with a term that is a further +restriction, you may remove that term. If a license document contains +a further restriction but permits relicensing or conveying under this +License, you may add to a covered work material governed by the terms +of that license document, provided that the further restriction does +not survive such relicensing or conveying. + + If you add terms to a covered work in accord with this section, you +must place, in the relevant source files, a statement of the +additional terms that apply to those files, or a notice indicating +where to find the applicable terms. + + Additional terms, permissive or non-permissive, may be stated in the +form of a separately written license, or stated as exceptions; +the above requirements apply either way. + + 8. Termination. + + You may not propagate or modify a covered work except as expressly +provided under this License. Any attempt otherwise to propagate or +modify it is void, and will automatically terminate your rights under +this License (including any patent licenses granted under the third +paragraph of section 11). + + However, if you cease all violation of this License, then your +license from a particular copyright holder is reinstated (a) +provisionally, unless and until the copyright holder explicitly and +finally terminates your license, and (b) permanently, if the copyright +holder fails to notify you of the violation by some reasonable means +prior to 60 days after the cessation. + + Moreover, your license from a particular copyright holder is +reinstated permanently if the copyright holder notifies you of the +violation by some reasonable means, this is the first time you have +received notice of violation of this License (for any work) from that +copyright holder, and you cure the violation prior to 30 days after +your receipt of the notice. + + Termination of your rights under this section does not terminate the +licenses of parties who have received copies or rights from you under +this License. If your rights have been terminated and not permanently +reinstated, you do not qualify to receive new licenses for the same +material under section 10. + + 9. Acceptance Not Required for Having Copies. + + You are not required to accept this License in order to receive or +run a copy of the Program. Ancillary propagation of a covered work +occurring solely as a consequence of using peer-to-peer transmission +to receive a copy likewise does not require acceptance. However, +nothing other than this License grants you permission to propagate or +modify any covered work. These actions infringe copyright if you do +not accept this License. Therefore, by modifying or propagating a +covered work, you indicate your acceptance of this License to do so. + + 10. Automatic Licensing of Downstream Recipients. + + Each time you convey a covered work, the recipient automatically +receives a license from the original licensors, to run, modify and +propagate that work, subject to this License. You are not responsible +for enforcing compliance by third parties with this License. + + An "entity transaction" is a transaction transferring control of an +organization, or substantially all assets of one, or subdividing an +organization, or merging organizations. If propagation of a covered +work results from an entity transaction, each party to that +transaction who receives a copy of the work also receives whatever +licenses to the work the party's predecessor in interest had or could +give under the previous paragraph, plus a right to possession of the +Corresponding Source of the work from the predecessor in interest, if +the predecessor has it or can get it with reasonable efforts. + + You may not impose any further restrictions on the exercise of the +rights granted or affirmed under this License. For example, you may +not impose a license fee, royalty, or other charge for exercise of +rights granted under this License, and you may not initiate litigation +(including a cross-claim or counterclaim in a lawsuit) alleging that +any patent claim is infringed by making, using, selling, offering for +sale, or importing the Program or any portion of it. + + 11. Patents. + + A "contributor" is a copyright holder who authorizes use under this +License of the Program or a work on which the Program is based. The +work thus licensed is called the contributor's "contributor version". + + A contributor's "essential patent claims" are all patent claims +owned or controlled by the contributor, whether already acquired or +hereafter acquired, that would be infringed by some manner, permitted +by this License, of making, using, or selling its contributor version, +but do not include claims that would be infringed only as a +consequence of further modification of the contributor version. For +purposes of this definition, "control" includes the right to grant +patent sublicenses in a manner consistent with the requirements of +this License. + + Each contributor grants you a non-exclusive, worldwide, royalty-free +patent license under the contributor's essential patent claims, to +make, use, sell, offer for sale, import and otherwise run, modify and +propagate the contents of its contributor version. + + In the following three paragraphs, a "patent license" is any express +agreement or commitment, however denominated, not to enforce a patent +(such as an express permission to practice a patent or covenant not to +sue for patent infringement). To "grant" such a patent license to a +party means to make such an agreement or commitment not to enforce a +patent against the party. + + If you convey a covered work, knowingly relying on a patent license, +and the Corresponding Source of the work is not available for anyone +to copy, free of charge and under the terms of this License, through a +publicly available network server or other readily accessible means, +then you must either (1) cause the Corresponding Source to be so +available, or (2) arrange to deprive yourself of the benefit of the +patent license for this particular work, or (3) arrange, in a manner +consistent with the requirements of this License, to extend the patent +license to downstream recipients. "Knowingly relying" means you have +actual knowledge that, but for the patent license, your conveying the +covered work in a country, or your recipient's use of the covered work +in a country, would infringe one or more identifiable patents in that +country that you have reason to believe are valid. + + If, pursuant to or in connection with a single transaction or +arrangement, you convey, or propagate by procuring conveyance of, a +covered work, and grant a patent license to some of the parties +receiving the covered work authorizing them to use, propagate, modify +or convey a specific copy of the covered work, then the patent license +you grant is automatically extended to all recipients of the covered +work and works based on it. + + A patent license is "discriminatory" if it does not include within +the scope of its coverage, prohibits the exercise of, or is +conditioned on the non-exercise of one or more of the rights that are +specifically granted under this License. You may not convey a covered +work if you are a party to an arrangement with a third party that is +in the business of distributing software, under which you make payment +to the third party based on the extent of your activity of conveying +the work, and under which the third party grants, to any of the +parties who would receive the covered work from you, a discriminatory +patent license (a) in connection with copies of the covered work +conveyed by you (or copies made from those copies), or (b) primarily +for and in connection with specific products or compilations that +contain the covered work, unless you entered into that arrangement, +or that patent license was granted, prior to 28 March 2007. + + Nothing in this License shall be construed as excluding or limiting +any implied license or other defenses to infringement that may +otherwise be available to you under applicable patent law. + + 12. No Surrender of Others' Freedom. + + If conditions are imposed on you (whether by court order, agreement or +otherwise) that contradict the conditions of this License, they do not +excuse you from the conditions of this License. If you cannot convey a +covered work so as to satisfy simultaneously your obligations under this +License and any other pertinent obligations, then as a consequence you may +not convey it at all. For example, if you agree to terms that obligate you +to collect a royalty for further conveying from those to whom you convey +the Program, the only way you could satisfy both those terms and this +License would be to refrain entirely from conveying the Program. + + 13. Use with the GNU Affero General Public License. + + Notwithstanding any other provision of this License, you have +permission to link or combine any covered work with a work licensed +under version 3 of the GNU Affero General Public License into a single +combined work, and to convey the resulting work. The terms of this +License will continue to apply to the part which is the covered work, +but the special requirements of the GNU Affero General Public License, +section 13, concerning interaction through a network will apply to the +combination as such. + + 14. Revised Versions of this License. + + The Free Software Foundation may publish revised and/or new versions of +the GNU General Public License from time to time. Such new versions will +be similar in spirit to the present version, but may differ in detail to +address new problems or concerns. + + Each version is given a distinguishing version number. If the +Program specifies that a certain numbered version of the GNU General +Public License "or any later version" applies to it, you have the +option of following the terms and conditions either of that numbered +version or of any later version published by the Free Software +Foundation. If the Program does not specify a version number of the +GNU General Public License, you may choose any version ever published +by the Free Software Foundation. + + If the Program specifies that a proxy can decide which future +versions of the GNU General Public License can be used, that proxy's +public statement of acceptance of a version permanently authorizes you +to choose that version for the Program. + + Later license versions may give you additional or different +permissions. However, no additional obligations are imposed on any +author or copyright holder as a result of your choosing to follow a +later version. + + 15. Disclaimer of Warranty. + + THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY +APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT +HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY +OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, +THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR +PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM +IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF +ALL NECESSARY SERVICING, REPAIR OR CORRECTION. + + 16. Limitation of Liability. + + IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING +WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS +THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY +GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE +USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF +DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD +PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), +EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF +SUCH DAMAGES. + + 17. Interpretation of Sections 15 and 16. + + If the disclaimer of warranty and limitation of liability provided +above cannot be given local legal effect according to their terms, +reviewing courts shall apply local law that most closely approximates +an absolute waiver of all civil liability in connection with the +Program, unless a warranty or assumption of liability accompanies a +copy of the Program in return for a fee. + + END OF TERMS AND CONDITIONS + + How to Apply These Terms to Your New Programs + + If you develop a new program, and you want it to be of the greatest +possible use to the public, the best way to achieve this is to make it +free software which everyone can redistribute and change under these terms. + + To do so, attach the following notices to the program. It is safest +to attach them to the start of each source file to most effectively +state the exclusion of warranty; and each file should have at least +the "copyright" line and a pointer to where the full notice is found. + + <one line to give the program's name and a brief idea of what it does.> + Copyright (C) <year> <name of author> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + +Also add information on how to contact you by electronic and paper mail. + + If the program does terminal interaction, make it output a short +notice like this when it starts in an interactive mode: + + <program> Copyright (C) <year> <name of author> + This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. + This is free software, and you are welcome to redistribute it + under certain conditions; type `show c' for details. + +The hypothetical commands `show w' and `show c' should show the appropriate +parts of the General Public License. Of course, your program's commands +might be different; for a GUI interface, you would use an "about box". + + You should also get your employer (if you work as a programmer) or school, +if any, to sign a "copyright disclaimer" for the program, if necessary. +For more information on this, and how to apply and follow the GNU GPL, see +<http://www.gnu.org/licenses/>. + + The GNU General Public License does not permit incorporating your program +into proprietary programs. If your program is a subroutine library, you +may consider it more useful to permit linking proprietary applications with +the library. If this is what you want to do, use the GNU Lesser General +Public License instead of this License. But first, please read +<http://www.gnu.org/philosophy/why-not-lgpl.html>. diff --git a/FEAST/FEAST/MIToolbox/COPYING.LESSER b/FEAST/FEAST/MIToolbox/COPYING.LESSER new file mode 100644 index 0000000..cca7fc2 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/COPYING.LESSER @@ -0,0 +1,165 @@ + GNU LESSER GENERAL PUBLIC LICENSE + Version 3, 29 June 2007 + + Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/> + Everyone is permitted to copy and distribute verbatim copies + of this license document, but changing it is not allowed. + + + This version of the GNU Lesser General Public License incorporates +the terms and conditions of version 3 of the GNU General Public +License, supplemented by the additional permissions listed below. + + 0. Additional Definitions. + + As used herein, "this License" refers to version 3 of the GNU Lesser +General Public License, and the "GNU GPL" refers to version 3 of the GNU +General Public License. + + "The Library" refers to a covered work governed by this License, +other than an Application or a Combined Work as defined below. + + An "Application" is any work that makes use of an interface provided +by the Library, but which is not otherwise based on the Library. +Defining a subclass of a class defined by the Library is deemed a mode +of using an interface provided by the Library. + + A "Combined Work" is a work produced by combining or linking an +Application with the Library. The particular version of the Library +with which the Combined Work was made is also called the "Linked +Version". + + The "Minimal Corresponding Source" for a Combined Work means the +Corresponding Source for the Combined Work, excluding any source code +for portions of the Combined Work that, considered in isolation, are +based on the Application, and not on the Linked Version. + + The "Corresponding Application Code" for a Combined Work means the +object code and/or source code for the Application, including any data +and utility programs needed for reproducing the Combined Work from the +Application, but excluding the System Libraries of the Combined Work. + + 1. Exception to Section 3 of the GNU GPL. + + You may convey a covered work under sections 3 and 4 of this License +without being bound by section 3 of the GNU GPL. + + 2. Conveying Modified Versions. + + If you modify a copy of the Library, and, in your modifications, a +facility refers to a function or data to be supplied by an Application +that uses the facility (other than as an argument passed when the +facility is invoked), then you may convey a copy of the modified +version: + + a) under this License, provided that you make a good faith effort to + ensure that, in the event an Application does not supply the + function or data, the facility still operates, and performs + whatever part of its purpose remains meaningful, or + + b) under the GNU GPL, with none of the additional permissions of + this License applicable to that copy. + + 3. Object Code Incorporating Material from Library Header Files. + + The object code form of an Application may incorporate material from +a header file that is part of the Library. You may convey such object +code under terms of your choice, provided that, if the incorporated +material is not limited to numerical parameters, data structure +layouts and accessors, or small macros, inline functions and templates +(ten or fewer lines in length), you do both of the following: + + a) Give prominent notice with each copy of the object code that the + Library is used in it and that the Library and its use are + covered by this License. + + b) Accompany the object code with a copy of the GNU GPL and this license + document. + + 4. Combined Works. + + You may convey a Combined Work under terms of your choice that, +taken together, effectively do not restrict modification of the +portions of the Library contained in the Combined Work and reverse +engineering for debugging such modifications, if you also do each of +the following: + + a) Give prominent notice with each copy of the Combined Work that + the Library is used in it and that the Library and its use are + covered by this License. + + b) Accompany the Combined Work with a copy of the GNU GPL and this license + document. + + c) For a Combined Work that displays copyright notices during + execution, include the copyright notice for the Library among + these notices, as well as a reference directing the user to the + copies of the GNU GPL and this license document. + + d) Do one of the following: + + 0) Convey the Minimal Corresponding Source under the terms of this + License, and the Corresponding Application Code in a form + suitable for, and under terms that permit, the user to + recombine or relink the Application with a modified version of + the Linked Version to produce a modified Combined Work, in the + manner specified by section 6 of the GNU GPL for conveying + Corresponding Source. + + 1) Use a suitable shared library mechanism for linking with the + Library. A suitable mechanism is one that (a) uses at run time + a copy of the Library already present on the user's computer + system, and (b) will operate properly with a modified version + of the Library that is interface-compatible with the Linked + Version. + + e) Provide Installation Information, but only if you would otherwise + be required to provide such information under section 6 of the + GNU GPL, and only to the extent that such information is + necessary to install and execute a modified version of the + Combined Work produced by recombining or relinking the + Application with a modified version of the Linked Version. (If + you use option 4d0, the Installation Information must accompany + the Minimal Corresponding Source and Corresponding Application + Code. If you use option 4d1, you must provide the Installation + Information in the manner specified by section 6 of the GNU GPL + for conveying Corresponding Source.) + + 5. Combined Libraries. + + You may place library facilities that are a work based on the +Library side by side in a single library together with other library +facilities that are not Applications and are not covered by this +License, and convey such a combined library under terms of your +choice, if you do both of the following: + + a) Accompany the combined library with a copy of the same work based + on the Library, uncombined with any other library facilities, + conveyed under the terms of this License. + + b) Give prominent notice with the combined library that part of it + is a work based on the Library, and explaining where to find the + accompanying uncombined form of the same work. + + 6. Revised Versions of the GNU Lesser General Public License. + + The Free Software Foundation may publish revised and/or new versions +of the GNU Lesser General Public License from time to time. Such new +versions will be similar in spirit to the present version, but may +differ in detail to address new problems or concerns. + + Each version is given a distinguishing version number. If the +Library as you received it specifies that a certain numbered version +of the GNU Lesser General Public License "or any later version" +applies to it, you have the option of following the terms and +conditions either of that published version or of any later version +published by the Free Software Foundation. If the Library as you +received it does not specify a version number of the GNU Lesser +General Public License, you may choose any version of the GNU Lesser +General Public License ever published by the Free Software Foundation. + + If the Library as you received it specifies that a proxy can decide +whether future versions of the GNU Lesser General Public License shall +apply, that proxy's public statement of acceptance of any version is +permanent authorization for you to choose that version for the +Library. diff --git a/FEAST/FEAST/MIToolbox/CalculateProbability.c b/FEAST/FEAST/MIToolbox/CalculateProbability.c new file mode 100644 index 0000000..6d4f19b --- /dev/null +++ b/FEAST/FEAST/MIToolbox/CalculateProbability.c @@ -0,0 +1,184 @@ +/******************************************************************************* +** CalculateProbability.cpp +** Part of the mutual information toolbox +** +** Contains functions to calculate the probability of each state in the array +** and to calculate the probability of the joint state of two arrays +** +** Author: Adam Pocock +** Created 17/2/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#include "MIToolbox.h" +#include "ArrayOperations.h" +#include "CalculateProbability.h" + +JointProbabilityState calculateJointProbability(double *firstVector, double *secondVector, int vectorLength) +{ + int *firstNormalisedVector; + int *secondNormalisedVector; + int *firstStateCounts; + int *secondStateCounts; + int *jointStateCounts; + double *firstStateProbs; + double *secondStateProbs; + double *jointStateProbs; + int firstNumStates; + int secondNumStates; + int jointNumStates; + int i; + double length = vectorLength; + JointProbabilityState state; + + firstNormalisedVector = (int *) CALLOC_FUNC(vectorLength,sizeof(int)); + secondNormalisedVector = (int *) CALLOC_FUNC(vectorLength,sizeof(int)); + + firstNumStates = normaliseArray(firstVector,firstNormalisedVector,vectorLength); + secondNumStates = normaliseArray(secondVector,secondNormalisedVector,vectorLength); + jointNumStates = firstNumStates * secondNumStates; + + firstStateCounts = (int *) CALLOC_FUNC(firstNumStates,sizeof(int)); + secondStateCounts = (int *) CALLOC_FUNC(secondNumStates,sizeof(int)); + jointStateCounts = (int *) CALLOC_FUNC(jointNumStates,sizeof(int)); + + firstStateProbs = (double *) CALLOC_FUNC(firstNumStates,sizeof(double)); + secondStateProbs = (double *) CALLOC_FUNC(secondNumStates,sizeof(double)); + jointStateProbs = (double *) CALLOC_FUNC(jointNumStates,sizeof(double)); + + /* optimised version, less numerically stable + double fractionalState = 1.0 / vectorLength; + + for (i = 0; i < vectorLength; i++) + { + firstStateProbs[firstNormalisedVector[i]] += fractionalState; + secondStateProbs[secondNormalisedVector[i]] += fractionalState; + jointStateProbs[secondNormalisedVector[i] * firstNumStates + firstNormalisedVector[i]] += fractionalState; + } + */ + + /* Optimised for number of FP operations now O(states) instead of O(vectorLength) */ + for (i = 0; i < vectorLength; i++) + { + firstStateCounts[firstNormalisedVector[i]] += 1; + secondStateCounts[secondNormalisedVector[i]] += 1; + jointStateCounts[secondNormalisedVector[i] * firstNumStates + firstNormalisedVector[i]] += 1; + } + + for (i = 0; i < firstNumStates; i++) + { + firstStateProbs[i] = firstStateCounts[i] / length; + } + + for (i = 0; i < secondNumStates; i++) + { + secondStateProbs[i] = secondStateCounts[i] / length; + } + + for (i = 0; i < jointNumStates; i++) + { + jointStateProbs[i] = jointStateCounts[i] / length; + } + + FREE_FUNC(firstNormalisedVector); + FREE_FUNC(secondNormalisedVector); + FREE_FUNC(firstStateCounts); + FREE_FUNC(secondStateCounts); + FREE_FUNC(jointStateCounts); + + firstNormalisedVector = NULL; + secondNormalisedVector = NULL; + firstStateCounts = NULL; + secondStateCounts = NULL; + jointStateCounts = NULL; + + /* + **typedef struct + **{ + ** double *jointProbabilityVector; + ** int numJointStates; + ** double *firstProbabilityVector; + ** int numFirstStates; + ** double *secondProbabilityVector; + ** int numSecondStates; + **} JointProbabilityState; + */ + + state.jointProbabilityVector = jointStateProbs; + state.numJointStates = jointNumStates; + state.firstProbabilityVector = firstStateProbs; + state.numFirstStates = firstNumStates; + state.secondProbabilityVector = secondStateProbs; + state.numSecondStates = secondNumStates; + + return state; +}/*calculateJointProbability(double *,double *, int)*/ + +ProbabilityState calculateProbability(double *dataVector, int vectorLength) +{ + int *normalisedVector; + int *stateCounts; + double *stateProbs; + int numStates; + /*double fractionalState;*/ + ProbabilityState state; + int i; + double length = vectorLength; + + normalisedVector = (int *) CALLOC_FUNC(vectorLength,sizeof(int)); + + numStates = normaliseArray(dataVector,normalisedVector,vectorLength); + + stateCounts = (int *) CALLOC_FUNC(numStates,sizeof(int)); + stateProbs = (double *) CALLOC_FUNC(numStates,sizeof(double)); + + /* optimised version, may have floating point problems + fractionalState = 1.0 / vectorLength; + + for (i = 0; i < vectorLength; i++) + { + stateProbs[normalisedVector[i]] += fractionalState; + } + */ + + /* Optimised for number of FP operations now O(states) instead of O(vectorLength) */ + for (i = 0; i < vectorLength; i++) + { + stateCounts[normalisedVector[i]] += 1; + } + + for (i = 0; i < numStates; i++) + { + stateProbs[i] = stateCounts[i] / length; + } + + FREE_FUNC(stateCounts); + FREE_FUNC(normalisedVector); + + stateCounts = NULL; + normalisedVector = NULL; + + state.probabilityVector = stateProbs; + state.numStates = numStates; + + return state; +}/*calculateProbability(double *,int)*/ + diff --git a/FEAST/FEAST/MIToolbox/CalculateProbability.h b/FEAST/FEAST/MIToolbox/CalculateProbability.h new file mode 100644 index 0000000..d5e9d3e --- /dev/null +++ b/FEAST/FEAST/MIToolbox/CalculateProbability.h @@ -0,0 +1,80 @@ +/******************************************************************************* +** CalculateProbability.h +** Part of the mutual information toolbox +** +** Contains functions to calculate the probability of each state in the array +** and to calculate the probability of the joint state of two arrays +** +** Author: Adam Pocock +** Created 17/2/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#ifndef __CalculateProbability_H +#define __CalculateProbability_H + +#ifdef __cplusplus +extern "C" { +#endif + +typedef struct jpState +{ + double *jointProbabilityVector; + int numJointStates; + double *firstProbabilityVector; + int numFirstStates; + double *secondProbabilityVector; + int numSecondStates; +} JointProbabilityState; + +typedef struct pState +{ + double *probabilityVector; + int numStates; +} ProbabilityState; + +/******************************************************************************* +** calculateJointProbability returns the joint probability vector of two vectors +** and the marginal probability vectors in a struct. +** It is the base operation for all information theory calculations involving +** two or more variables. +** +** length(firstVector) == length(secondVector) == vectorLength otherwise there +** will be a segmentation fault +*******************************************************************************/ +JointProbabilityState calculateJointProbability(double *firstVector, double *secondVector, int vectorLength); + +/******************************************************************************* +** calculateProbability returns the probability vector from one vector. +** It is the base operation for all information theory calculations involving +** one variable +** +** length(dataVector) == vectorLength otherwise there +** will be a segmentation fault +*******************************************************************************/ +ProbabilityState calculateProbability(double *dataVector, int vectorLength); + +#ifdef __cplusplus +} +#endif + +#endif + diff --git a/FEAST/FEAST/MIToolbox/CompileMIToolbox.m b/FEAST/FEAST/MIToolbox/CompileMIToolbox.m new file mode 100644 index 0000000..a8d9e92 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/CompileMIToolbox.m @@ -0,0 +1,4 @@ +% Compiles the MIToolbox functions + +mex MIToolboxMex.c MutualInformation.c Entropy.c CalculateProbability.c ArrayOperations.c +mex RenyiMIToolboxMex.c RenyiMutualInformation.c RenyiEntropy.c CalculateProbability.c ArrayOperations.c diff --git a/FEAST/FEAST/MIToolbox/Entropy.c b/FEAST/FEAST/MIToolbox/Entropy.c new file mode 100644 index 0000000..3f37cc1 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/Entropy.c @@ -0,0 +1,130 @@ +/******************************************************************************* +** Entropy.cpp +** Part of the mutual information toolbox +** +** Contains functions to calculate the entropy of a single variable H(X), +** the joint entropy of two variables H(X,Y), and the conditional entropy +** H(X|Y) +** +** Author: Adam Pocock +** Created 19/2/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#include "MIToolbox.h" +#include "CalculateProbability.h" +#include "Entropy.h" + +double calculateEntropy(double *dataVector, int vectorLength) +{ + double entropy = 0.0; + double tempValue = 0.0; + int i; + ProbabilityState state = calculateProbability(dataVector,vectorLength); + + /*H(X) = - sum p(x) log p(x)*/ + for (i = 0; i < state.numStates; i++) + { + tempValue = state.probabilityVector[i]; + + if (tempValue > 0) + { + entropy -= tempValue * log(tempValue); + } + } + + entropy /= log(2.0); + + FREE_FUNC(state.probabilityVector); + state.probabilityVector = NULL; + + return entropy; +}/*calculateEntropy(double *,int)*/ + +double calculateJointEntropy(double *firstVector, double *secondVector, int vectorLength) +{ + double jointEntropy = 0.0; + double tempValue = 0.0; + int i; + JointProbabilityState state = calculateJointProbability(firstVector,secondVector,vectorLength); + + /*H(XY) = - sumx sumy p(xy) log p(xy)*/ + for (i = 0; i < state.numJointStates; i++) + { + tempValue = state.jointProbabilityVector[i]; + if (tempValue > 0) + { + jointEntropy -= tempValue * log(tempValue); + } + } + + jointEntropy /= log(2.0); + + FREE_FUNC(state.firstProbabilityVector); + state.firstProbabilityVector = NULL; + FREE_FUNC(state.secondProbabilityVector); + state.secondProbabilityVector = NULL; + FREE_FUNC(state.jointProbabilityVector); + state.jointProbabilityVector = NULL; + + return jointEntropy; +}/*calculateJointEntropy(double *, double *, int)*/ + +double calculateConditionalEntropy(double *dataVector, double *conditionVector, int vectorLength) +{ + /* + ** Conditional entropy + ** H(X|Y) = - sumx sumy p(xy) log p(xy)/p(y) + */ + + double condEntropy = 0.0; + double jointValue = 0.0; + double condValue = 0.0; + int i; + JointProbabilityState state = calculateJointProbability(dataVector,conditionVector,vectorLength); + + /*H(X|Y) = - sumx sumy p(xy) log p(xy)/p(y)*/ + /* to index by numFirstStates use modulus of i + ** to index by numSecondStates use integer division of i by numFirstStates + */ + for (i = 0; i < state.numJointStates; i++) + { + jointValue = state.jointProbabilityVector[i]; + condValue = state.secondProbabilityVector[i / state.numFirstStates]; + if ((jointValue > 0) && (condValue > 0)) + { + condEntropy -= jointValue * log(jointValue / condValue); + } + } + + condEntropy /= log(2.0); + + FREE_FUNC(state.firstProbabilityVector); + state.firstProbabilityVector = NULL; + FREE_FUNC(state.secondProbabilityVector); + state.secondProbabilityVector = NULL; + FREE_FUNC(state.jointProbabilityVector); + state.jointProbabilityVector = NULL; + + return condEntropy; + +}/*calculateConditionalEntropy(double *, double *, int)*/ + diff --git a/FEAST/FEAST/MIToolbox/Entropy.h b/FEAST/FEAST/MIToolbox/Entropy.h new file mode 100644 index 0000000..4bdd697 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/Entropy.h @@ -0,0 +1,71 @@ +/******************************************************************************* +** Entropy.h +** Part of the mutual information toolbox +** +** Contains functions to calculate the entropy of a single variable H(X), +** the joint entropy of two variables H(X,Y), and the conditional entropy +** H(X|Y) +** +** Author: Adam Pocock +** Created 19/2/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#ifndef __Entropy_H +#define __Entropy_H + +#ifdef __cplusplus +extern "C" { +#endif + +/******************************************************************************* +** calculateEntropy returns the entropy in log base 2 of dataVector +** H(X) +** +** length(dataVector) == vectorLength otherwise there +** will be a segmentation fault +*******************************************************************************/ +double calculateEntropy(double *dataVector, int vectorLength); + +/******************************************************************************* +** calculateJointEntropy returns the entropy in log base 2 of the joint +** variable of firstVector and secondVector H(XY) +** +** length(firstVector) == length(secondVector) == vectorLength otherwise there +** will be a segmentation fault +*******************************************************************************/ +double calculateJointEntropy(double *firstVector, double *secondVector, int vectorLength); + +/******************************************************************************* +** calculateConditionalEntropy returns the entropy in log base 2 of dataVector +** conditioned on conditionVector, H(X|Y) +** +** length(dataVector) == length(conditionVector) == vectorLength otherwise there +** will be a segmentation fault +*******************************************************************************/ +double calculateConditionalEntropy(double *dataVector, double *conditionVector, int vectorLength); + +#ifdef __cplusplus +} +#endif + +#endif + diff --git a/FEAST/FEAST/MIToolbox/MIToolbox.h b/FEAST/FEAST/MIToolbox/MIToolbox.h new file mode 100644 index 0000000..bba6284 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/MIToolbox.h @@ -0,0 +1,53 @@ +/******************************************************************************* +** +** MIToolbox.h +** Provides the header files and #defines to ensure compatibility with MATLAB +** and C/C++. Uncomment the correct lines to setup the correct memory +** allocation and freeing operations. +** +** Author: Adam Pocock +** Created 17/2/2010 +** +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#ifndef __MIToolbox_H +#define __MIToolbox_H + +#include <math.h> +#include <string.h> + +#ifdef COMPILE_C + #define C_IMPLEMENTATION + #include <stdio.h> + #include <stdlib.h> + #define CALLOC_FUNC calloc + #define FREE_FUNC free +#else + #define MEX_IMPLEMENTATION + #include "mex.h" + #define CALLOC_FUNC mxCalloc + #define FREE_FUNC mxFree + #define printf mexPrintf /*for Octave-3.2*/ +#endif + +#endif + diff --git a/FEAST/FEAST/MIToolbox/MIToolbox.m b/FEAST/FEAST/MIToolbox/MIToolbox.m new file mode 100644 index 0000000..880924a --- /dev/null +++ b/FEAST/FEAST/MIToolbox/MIToolbox.m @@ -0,0 +1,83 @@ +function [varargout] = MIToolbox(functionName, varargin) +%function [varargout] = MIToolbox(functionName, varargin) +% +%Provides access to the functions in MIToolboxMex +% +%Expects column vectors, will not work with row vectors +% +%Function list +%"joint" = joint variable of the matrix +%"entropy" or "h" = H(X) +%"ConditionalEntropy" or "condh" = H(X|Y) +%"mi" = I(X;Y) +%"ConditionalMI" or "cmi" = I(X;Y|Z) +% +%Arguments and returned values +%[jointVariable] = joint(matrix) +%[entropy] = H(X) = H(vector) +%[entropy] = H(X|Y) = H(vector,condition) +%[mi] = I(X;Y) = I(vector,target) +%[mi] = I(X;Y|Z) = I(vector,target,condition) +% +%Internal MIToolbox function number +%Joint = 3 +%Entropy = 4 +%Conditional Entropy = 6 +%Mutual Information = 7 +%Conditional MI = 8 + +if (strcmpi(functionName,'Joint') || strcmpi(functionName,'Merge')) + [varargout{1}] = MIToolboxMex(3,varargin{1}); +elseif (strcmpi(functionName,'Entropy') || strcmpi(functionName,'h')) + %disp('Calculating Entropy'); + if (size(varargin{1},2)>1) + mergedVector = MIToolboxMex(3,varargin{1}); + else + mergedVector = varargin{1}; + end + [varargout{1}] = MIToolboxMex(4,mergedVector); +elseif ((strcmpi(functionName,'ConditionalEntropy')) || strcmpi(functionName,'condh')) + if (size(varargin{1},2)>1) + mergedFirst = MIToolboxMex(3,varargin{1}); + else + mergedFirst = varargin{1}; + end + if (size(varargin{2},2)>1) + mergedSecond = MIToolboxMex(3,varargin{2}); + else + mergedSecond = varargin{2}; + end + [varargout{1}] = MIToolboxMex(6,mergedFirst,mergedSecond); +elseif (strcmpi(functionName,'mi')) + if (size(varargin{1},2)>1) + mergedFirst = MIToolboxMex(3,varargin{1}); + else + mergedFirst = varargin{1}; + end + if (size(varargin{2},2)>1) + mergedSecond = MIToolboxMex(3,varargin{2}); + else + mergedSecond = varargin{2}; + end + [varargout{1}] = MIToolboxMex(7,mergedFirst,mergedSecond); +elseif (strcmpi(functionName,'ConditionalMI') || strcmpi(functionName,'cmi')) + if (size(varargin{1},2)>1) + mergedFirst = MIToolboxMex(3,varargin{1}); + else + mergedFirst = varargin{1}; + end + if (size(varargin{2},2)>1) + mergedSecond = MIToolboxMex(3,varargin{2}); + else + mergedSecond = varargin{2}; + end + if (size(varargin{3},2)>1) + mergedThird = MIToolboxMex(3,varargin{3}); + else + mergedThird = varargin{3}; + end + [varargout{1}] = MIToolboxMex(8,mergedFirst,mergedSecond,mergedThird); +else + varargout{1} = 0; + disp(['Unrecognised functionName ' functionName]); +end diff --git a/FEAST/FEAST/MIToolbox/MIToolboxMex.c b/FEAST/FEAST/MIToolbox/MIToolboxMex.c new file mode 100644 index 0000000..e27009f --- /dev/null +++ b/FEAST/FEAST/MIToolbox/MIToolboxMex.c @@ -0,0 +1,494 @@ +/******************************************************************************* +** +** MIToolboxMex.cpp +** is the MATLAB entry point for the MIToolbox functions when called from +** a MATLAB/OCTAVE script. +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ +#include "MIToolbox.h" +#include "ArrayOperations.h" +#include "CalculateProbability.h" +#include "Entropy.h" +#include "MutualInformation.h" + +/******************************************************************************* +**entry point for the mex call +**nlhs - number of outputs +**plhs - pointer to array of outputs +**nrhs - number of inputs +**prhs - pointer to array of inputs +*******************************************************************************/ +void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) +{ + /***************************************************************************** + ** this function takes a flag and a variable number of arguments + ** depending on the value of the flag and returns either a construct + ** containing probability estimates, a merged vector or a double value + ** representing an entropy or mutual information + *****************************************************************************/ + + int flag, i, numberOfSamples, checkSamples, thirdCheckSamples, numberOfFeatures, checkFeatures, thirdCheckFeatures; + int numArities, errorTest; + double *dataVector, *condVector, *targetVector, *firstVector, *secondVector, *output, *numStates; + double *matrix, *mergedVector, *arities; + int *outputIntVector, *intArities; + + double *jointOutput, *numJointStates, *firstOutput, *numFirstStates, *secondOutput, *numSecondStates; + + ProbabilityState state; + JointProbabilityState jointState; + + /*if (nlhs != 1) + { + printf("Incorrect number of output arguments\n"); + }//if not 1 output + */ + if (nrhs == 2) + { + /*printf("Must be H(X), calculateProbability(X), merge(X), normaliseArray(X)\n");*/ + } + else if (nrhs == 3) + { + /*printf("Must be H(XY), H(X|Y), calculateJointProbability(XY), I(X;Y)\n");*/ + } + else if (nrhs == 4) + { + /*printf("Must be I(X;Y|Z)\n");*/ + } + else + { + printf("Incorrect number of arguments, format is MIToolbox(\"FLAG\",varargin)\n"); + } + + /* number to function map + ** 1 = calculateProbability + ** 2 = calculateJointProbability + ** 3 = mergeArrays + ** 4 = H(X) + ** 5 = H(XY) + ** 6 = H(X|Y) + ** 7 = I(X;Y) + ** 8 = I(X;Y|Z) + ** 9 = normaliseArray + */ + + flag = *mxGetPr(prhs[0]); + + switch (flag) + { + case 1: + { + /* + **calculateProbability + */ + numberOfSamples = mxGetM(prhs[1]); + dataVector = (double *) mxGetPr(prhs[1]); + + /*ProbabilityState calculateProbability(double *dataVector, int vectorLength);*/ + state = calculateProbability(dataVector,numberOfSamples); + + plhs[0] = mxCreateDoubleMatrix(state.numStates,1,mxREAL); + plhs[1] = mxCreateDoubleMatrix(1,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + numStates = (double *) mxGetPr(plhs[1]); + + *numStates = state.numStates; + + for (i = 0; i < state.numStates; i++) + { + output[i] = state.probabilityVector[i]; + } + + break; + }/*case 1 - calculateProbability*/ + case 2: + { + /* + **calculateJointProbability + */ + numberOfSamples = mxGetM(prhs[1]); + firstVector = (double *) mxGetPr(prhs[1]); + secondVector = (double *) mxGetPr(prhs[2]); + + /*JointProbabilityState calculateJointProbability(double *firstVector, double *secondVector int vectorLength);*/ + jointState = calculateJointProbability(firstVector,secondVector,numberOfSamples); + + plhs[0] = mxCreateDoubleMatrix(jointState.numJointStates,1,mxREAL); + plhs[1] = mxCreateDoubleMatrix(1,1,mxREAL); + plhs[2] = mxCreateDoubleMatrix(jointState.numFirstStates,1,mxREAL); + plhs[3] = mxCreateDoubleMatrix(1,1,mxREAL); + plhs[4] = mxCreateDoubleMatrix(jointState.numSecondStates,1,mxREAL); + plhs[5] = mxCreateDoubleMatrix(1,1,mxREAL); + jointOutput = (double *)mxGetPr(plhs[0]); + numJointStates = (double *) mxGetPr(plhs[1]); + firstOutput = (double *)mxGetPr(plhs[2]); + numFirstStates = (double *) mxGetPr(plhs[3]); + secondOutput = (double *)mxGetPr(plhs[4]); + numSecondStates = (double *) mxGetPr(plhs[5]); + + *numJointStates = jointState.numJointStates; + *numFirstStates = jointState.numFirstStates; + *numSecondStates = jointState.numSecondStates; + + for (i = 0; i < jointState.numJointStates; i++) + { + jointOutput[i] = jointState.jointProbabilityVector[i]; + } + for (i = 0; i < jointState.numFirstStates; i++) + { + firstOutput[i] = jointState.firstProbabilityVector[i]; + } + for (i = 0; i < jointState.numSecondStates; i++) + { + secondOutput[i] = jointState.secondProbabilityVector[i]; + } + + break; + }/*case 2 - calculateJointProbability */ + case 3: + { + /* + **mergeArrays + */ + numberOfSamples = mxGetM(prhs[1]); + numberOfFeatures = mxGetN(prhs[1]); + + numArities = 0; + if (nrhs > 2) + { + numArities = mxGetN(prhs[2]); + /*printf("arities = %d, features = %d, samples = %d\n",numArities,numberOfFeatures,numberOfSamples);*/ + } + + plhs[0] = mxCreateDoubleMatrix(0,0,mxREAL); + + if (numArities == 0) + { + /* + **no arities therefore compress output + */ + if ((numberOfFeatures > 0) && (numberOfSamples > 0)) + { + matrix = (double *) mxGetPr(prhs[1]); + mergedVector = (double *) mxCalloc(numberOfSamples,sizeof(double)); + + plhs[0] = mxCreateDoubleMatrix(numberOfSamples,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /*int mergeMultipleArrays(double *inputMatrix, double *outputVector, int matrixWidth, int vectorLength)*/ + mergeMultipleArrays(matrix, mergedVector, numberOfFeatures, numberOfSamples); + for (i = 0; i < numberOfSamples; i++) + { + output[i] = mergedVector[i]; + } + + mxFree(mergedVector); + mergedVector = NULL; + } + } + else if (numArities == numberOfFeatures) + { + if ((numberOfFeatures > 0) && (numberOfSamples > 0)) + { + + matrix = (double *) mxGetPr(prhs[1]); + mergedVector = (double *) mxCalloc(numberOfSamples,sizeof(double)); + + arities = (double *) mxGetPr(prhs[2]); + intArities = (int *) mxCalloc(numberOfFeatures,sizeof(int)); + for (i = 0; i < numArities; i++) + { + intArities[i] = (int) floor(arities[i]); + } + + /*int mergeMultipleArrays(double *inputMatrix, double *outputVector, int matrixWidth, int *arities, int vectorLength);*/ + errorTest = mergeMultipleArraysArities(matrix, mergedVector, numberOfFeatures, intArities, numberOfSamples); + + if (errorTest != -1) + { + plhs[0] = mxCreateDoubleMatrix(numberOfSamples,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + for (i = 0; i < numberOfSamples; i++) + { + output[i] = mergedVector[i]; + } + } + else + { + printf("Incorrect arities supplied. More states in data than specified\n"); + } + + mxFree(mergedVector); + mergedVector = NULL; + } + } + else + { + printf("Number of arities does not match number of features, arities should be a row vector\n"); + } + + break; + }/*case 3 - mergeArrays*/ + case 4: + { + /* + **H(X) + */ + numberOfSamples = mxGetM(prhs[1]); + numberOfFeatures = mxGetN(prhs[1]); + + dataVector = (double *) mxGetPr(prhs[1]); + + plhs[0] = mxCreateDoubleMatrix(1,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + if (numberOfFeatures == 1) + { + /*double calculateEntropy(double *dataVector, int vectorLength);*/ + *output = calculateEntropy(dataVector,numberOfSamples); + } + else + { + printf("No columns in input\n"); + *output = -1.0; + } + + break; + }/*case 4 - H(X)*/ + case 5: + { + /* + **H(XY) + */ + numberOfSamples = mxGetM(prhs[1]); + checkSamples = mxGetM(prhs[2]); + + numberOfFeatures = mxGetN(prhs[1]); + checkFeatures = mxGetN(prhs[2]); + + firstVector = mxGetPr(prhs[1]); + secondVector = mxGetPr(prhs[2]); + + plhs[0] = mxCreateDoubleMatrix(1,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + + if ((numberOfFeatures == 1) && (checkFeatures == 1)) + { + if ((numberOfSamples == 0) && (checkSamples == 0)) + { + *output = 0.0; + } + else if (numberOfSamples == 0) + { + *output = calculateEntropy(secondVector,numberOfSamples); + } + else if (checkSamples == 0) + { + *output = calculateEntropy(firstVector,numberOfSamples); + } + else if (numberOfSamples == checkSamples) + { + /*double calculateJointEntropy(double *firstVector, double *secondVector, int vectorLength);*/ + *output = calculateJointEntropy(firstVector,secondVector,numberOfSamples); + } + else + { + printf("Vector lengths do not match, they must be the same length\n"); + *output = -1.0; + } + } + else + { + printf("No columns in input\n"); + *output = -1.0; + } + + break; + }/*case 5 - H(XY)*/ + case 6: + { + /* + **H(X|Y) + */ + numberOfSamples = mxGetM(prhs[1]); + checkSamples = mxGetM(prhs[2]); + + numberOfFeatures = mxGetN(prhs[1]); + checkFeatures = mxGetN(prhs[2]); + + dataVector = mxGetPr(prhs[1]); + condVector = mxGetPr(prhs[2]); + + plhs[0] = mxCreateDoubleMatrix(1,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + if ((numberOfFeatures == 1) && (checkFeatures == 1)) + { + if (numberOfSamples == 0) + { + *output = 0.0; + } + else if (checkSamples == 0) + { + *output = calculateEntropy(dataVector,numberOfSamples); + } + else if (numberOfSamples == checkSamples) + { + /*double calculateConditionalEntropy(double *dataVector, double *condVector, int vectorLength);*/ + *output = calculateConditionalEntropy(dataVector,condVector,numberOfSamples); + } + else + { + printf("Vector lengths do not match, they must be the same length\n"); + *output = -1.0; + } + } + else + { + printf("No columns in input\n"); + *output = -1.0; + } + break; + }/*case 6 - H(X|Y)*/ + case 7: + { + /* + **I(X;Y) + */ + numberOfSamples = mxGetM(prhs[1]); + checkSamples = mxGetM(prhs[2]); + + numberOfFeatures = mxGetN(prhs[1]); + checkFeatures = mxGetN(prhs[2]); + + firstVector = mxGetPr(prhs[1]); + secondVector = mxGetPr(prhs[2]); + + plhs[0] = mxCreateDoubleMatrix(1,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + if ((numberOfFeatures == 1) && (checkFeatures == 1)) + { + if ((numberOfSamples == 0) || (checkSamples == 0)) + { + *output = 0.0; + } + else if (numberOfSamples == checkSamples) + { + /*double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength);*/ + *output = calculateMutualInformation(firstVector,secondVector,numberOfSamples); + } + else + { + printf("Vector lengths do not match, they must be the same length\n"); + *output = -1.0; + } + } + else + { + printf("No columns in input\n"); + *output = -1.0; + } + break; + }/*case 7 - I(X;Y)*/ + case 8: + { + /* + **I(X;Y|Z) + */ + numberOfSamples = mxGetM(prhs[1]); + checkSamples = mxGetM(prhs[2]); + thirdCheckSamples = mxGetM(prhs[3]); + + numberOfFeatures = mxGetN(prhs[1]); + checkFeatures = mxGetN(prhs[2]); + thirdCheckFeatures = mxGetN(prhs[3]); + + firstVector = mxGetPr(prhs[1]); + targetVector = mxGetPr(prhs[2]); + condVector = mxGetPr(prhs[3]); + + plhs[0] = mxCreateDoubleMatrix(1,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + if ((numberOfFeatures == 1) && (checkFeatures == 1)) + { + if ((numberOfSamples == 0) || (checkSamples == 0)) + { + *output = 0.0; + } + else if ((thirdCheckSamples == 0) || (thirdCheckFeatures != 1)) + { + *output = calculateMutualInformation(firstVector,targetVector,numberOfSamples); + } + else if ((numberOfSamples == checkSamples) && (numberOfSamples == thirdCheckSamples)) + { + /*double calculateConditionalMutualInformation(double *firstVector, double *targetVector, double *condVector, int vectorLength);*/ + *output = calculateConditionalMutualInformation(firstVector,targetVector,condVector,numberOfSamples); + } + else + { + printf("Vector lengths do not match, they must be the same length\n"); + *output = -1.0; + } + } + else + { + printf("No columns in input\n"); + *output = -1.0; + } + break; + }/*case 8 - I(X;Y|Z)*/ + case 9: + { + /* + **normaliseArray + */ + numberOfSamples = mxGetM(prhs[1]); + dataVector = (double *) mxGetPr(prhs[1]); + + outputIntVector = (int *) mxCalloc(numberOfSamples,sizeof(int)); + + plhs[0] = mxCreateDoubleMatrix(numberOfSamples,1,mxREAL); + plhs[1] = mxCreateDoubleMatrix(1,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + numStates = (double *) mxGetPr(plhs[1]); + + /*int normaliseArray(double *inputVector, int *outputVector, int vectorLength);*/ + *numStates = normaliseArray(dataVector, outputIntVector, numberOfSamples); + + for (i = 0; i < numberOfSamples; i++) + { + output[i] = outputIntVector[i]; + } + + break; + }/*case 9 - normaliseArray*/ + default: + { + printf("Unrecognised flag\n"); + break; + }/*default*/ + }/*switch(flag)*/ + + return; +}/*mexFunction()*/ diff --git a/FEAST/FEAST/MIToolbox/MIToolboxMex.mexa64 b/FEAST/FEAST/MIToolbox/MIToolboxMex.mexa64 Binary files differnew file mode 100755 index 0000000..440611d --- /dev/null +++ b/FEAST/FEAST/MIToolbox/MIToolboxMex.mexa64 diff --git a/FEAST/FEAST/MIToolbox/Makefile b/FEAST/FEAST/MIToolbox/Makefile new file mode 100644 index 0000000..0135410 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/Makefile @@ -0,0 +1,84 @@ +#makefile for MIToolbox +#Author: Adam Pocock, apocock@cs.man.ac.uk +#Created 11/3/2010 +# +# +#Copyright 2010 Adam Pocock, The University Of Manchester +#www.cs.manchester.ac.uk +# +#This file is part of MIToolbox. +# +#MIToolbox is free software: you can redistribute it and/or modify +#it under the terms of the GNU Lesser General Public License as published by +#the Free Software Foundation, either version 3 of the License, or +#(at your option) any later version. +# +#MIToolbox is distributed in the hope that it will be useful, +#but WITHOUT ANY WARRANTY; without even the implied warranty of +#MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +#GNU Lesser General Public License for more details. +# +#You should have received a copy of the GNU Lesser General Public License +#along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. + +CXXFLAGS = -O3 -fPIC +COMPILER = gcc +objects = ArrayOperations.o CalculateProbability.o Entropy.o \ + MutualInformation.o RenyiEntropy.o RenyiMutualInformation.o + +libMIToolbox.so : $(objects) + $(COMPILER) $(CXXFLAGS) -shared -o libMIToolbox.so $(objects) + +RenyiMutualInformation.o: RenyiMutualInformation.c MIToolbox.h ArrayOperations.h \ + CalculateProbability.h RenyiEntropy.h + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c RenyiMutualInformation.c + +RenyiEntropy.o: RenyiEntropy.c MIToolbox.h ArrayOperations.h \ + CalculateProbability.h + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c RenyiEntropy.c + +MutualInformation.o: MutualInformation.c MIToolbox.h ArrayOperations.h \ + CalculateProbability.h Entropy.h MutualInformation.h + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c MutualInformation.c + +Entropy.o: Entropy.c MIToolbox.h ArrayOperations.h CalculateProbability.h \ + Entropy.h + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c Entropy.c + +CalculateProbability.o: CalculateProbability.c MIToolbox.h ArrayOperations.h \ + CalculateProbability.h + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c CalculateProbability.c + +ArrayOperations.o: ArrayOperations.c MIToolbox.h ArrayOperations.h + $(COMPILER) $(CXXFLAGS) -DCOMPILE_C -c ArrayOperations.c + +.PHONY : debug +debug: + $(MAKE) libMIToolbox.so "CXXFLAGS = -g -DDEBUG -fPIC" + +.PHONY : x86 +x86: + $(MAKE) libMIToolbox.so "CXXFLAGS = -O3 -fPIC -m32" + +.PHONY : x64 +x64: + $(MAKE) libMIToolbox.so "CXXFLAGS = -O3 -fPIC -m64" + +.PHONY : matlab +matlab: + mex MIToolboxMex.c MutualInformation.c Entropy.c CalculateProbability.c ArrayOperations.c + mex RenyiMIToolboxMex.c RenyiMutualInformation.c RenyiEntropy.c CalculateProbability.c ArrayOperations.c + +.PHONY : matlab-debug +matlab-debug: + mex -g MIToolboxMex.c MutualInformation.c Entropy.c CalculateProbability.c ArrayOperations.c + mex -g RenyiMIToolboxMex.c RenyiMutualInformation.c RenyiEntropy.c CalculateProbability.c ArrayOperations.c + +.PHONY : intel +intel: + $(MAKE) libMIToolbox.so "COMPILER = icc" "CXXFLAGS = -O2 -fPIC -xHost" + +.PHONY : clean +clean: + rm *.o libMIToolbox.so + diff --git a/FEAST/FEAST/MIToolbox/MutualInformation.c b/FEAST/FEAST/MIToolbox/MutualInformation.c new file mode 100644 index 0000000..0fb4766 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/MutualInformation.c @@ -0,0 +1,95 @@ +/******************************************************************************* +** MutualInformation.cpp +** Part of the mutual information toolbox +** +** Contains functions to calculate the mutual information of +** two variables X and Y, I(X;Y), to calculate the joint mutual information +** of two variables X & Z on the variable Y, I(XZ;Y), and the conditional +** mutual information I(x;Y|Z) +** +** Author: Adam Pocock +** Created 19/2/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#include "MIToolbox.h" +#include "ArrayOperations.h" +#include "CalculateProbability.h" +#include "Entropy.h" +#include "MutualInformation.h" + +double calculateMutualInformation(double *dataVector, double *targetVector, int vectorLength) +{ + double mutualInformation = 0.0; + int firstIndex,secondIndex; + int i; + JointProbabilityState state = calculateJointProbability(dataVector,targetVector,vectorLength); + + /* + ** I(X;Y) = sum sum p(xy) * log (p(xy)/p(x)p(y)) + */ + for (i = 0; i < state.numJointStates; i++) + { + firstIndex = i % state.numFirstStates; + secondIndex = i / state.numFirstStates; + + if ((state.jointProbabilityVector[i] > 0) && (state.firstProbabilityVector[firstIndex] > 0) && (state.secondProbabilityVector[secondIndex] > 0)) + { + /*double division is probably more stable than multiplying two small numbers together + ** mutualInformation += state.jointProbabilityVector[i] * log(state.jointProbabilityVector[i] / (state.firstProbabilityVector[firstIndex] * state.secondProbabilityVector[secondIndex])); + */ + mutualInformation += state.jointProbabilityVector[i] * log(state.jointProbabilityVector[i] / state.firstProbabilityVector[firstIndex] / state.secondProbabilityVector[secondIndex]); + } + } + + mutualInformation /= log(2.0); + + FREE_FUNC(state.firstProbabilityVector); + state.firstProbabilityVector = NULL; + FREE_FUNC(state.secondProbabilityVector); + state.secondProbabilityVector = NULL; + FREE_FUNC(state.jointProbabilityVector); + state.jointProbabilityVector = NULL; + + return mutualInformation; +}/*calculateMutualInformation(double *,double *,int)*/ + +double calculateConditionalMutualInformation(double *dataVector, double *targetVector, double *conditionVector, int vectorLength) +{ + double mutualInformation = 0.0; + double firstCondition, secondCondition; + double *mergedVector = (double *) CALLOC_FUNC(vectorLength,sizeof(double)); + + mergeArrays(targetVector,conditionVector,mergedVector,vectorLength); + + /* I(X;Y|Z) = H(X|Z) - H(X|YZ) */ + /* double calculateConditionalEntropy(double *dataVector, double *conditionVector, int vectorLength); */ + firstCondition = calculateConditionalEntropy(dataVector,conditionVector,vectorLength); + secondCondition = calculateConditionalEntropy(dataVector,mergedVector,vectorLength); + + mutualInformation = firstCondition - secondCondition; + + FREE_FUNC(mergedVector); + mergedVector = NULL; + + return mutualInformation; +}/*calculateConditionalMutualInformation(double *,double *,double *,int)*/ + diff --git a/FEAST/FEAST/MIToolbox/MutualInformation.h b/FEAST/FEAST/MIToolbox/MutualInformation.h new file mode 100644 index 0000000..1045912 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/MutualInformation.h @@ -0,0 +1,64 @@ +/******************************************************************************* +** MutualInformation.h +** Part of the mutual information toolbox +** +** Contains functions to calculate the mutual information of +** two variables X and Y, I(X;Y), to calculate the joint mutual information +** of two variables X & Z on the variable Y, I(XZ;Y), and the conditional +** mutual information I(x;Y|Z) +** +** Author: Adam Pocock +** Created 19/2/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#ifndef __MutualInformation_H +#define __MutualInformation_H + +#ifdef __cplusplus +extern "C" { +#endif + +/******************************************************************************* +** calculateMutualInformation returns the log base 2 mutual information between +** dataVector and targetVector, I(X;Y) +** +** length(dataVector) == length(targetVector) == vectorLength otherwise there +** will be a segmentation fault +*******************************************************************************/ +double calculateMutualInformation(double *dataVector, double *targetVector, int vectorLength); + +/******************************************************************************* +** calculateConditionalMutualInformation returns the log base 2 +** mutual information between dataVector and targetVector, conditioned on +** conditionVector, I(X;Y|Z) +** +** length(dataVector) == length(targetVector) == length(condtionVector) == vectorLength +** otherwise it will error with a segmentation fault +*******************************************************************************/ +double calculateConditionalMutualInformation(double *dataVector, double *targetVector, double *conditionVector, int vectorLength); + +#ifdef __cplusplus +} +#endif + +#endif + diff --git a/FEAST/FEAST/MIToolbox/README b/FEAST/FEAST/MIToolbox/README new file mode 100644 index 0000000..7abe43d --- /dev/null +++ b/FEAST/FEAST/MIToolbox/README @@ -0,0 +1,71 @@ +MIToolbox v1.03 for C/C++ and MATLAB/OCTAVE + +The MIToolbox contains a set of functions to calculate information theoretic +quantities from data, such as the entropy and mutual information. The toolbox +contains implementations of the most popular Shannon entropies, and also the +lesser known Renyi entropy. 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 rather +than communication channels and so expect all the data to be available before +execution and sample their own probability distributions from the data. + +Things you can do: + - 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 + +Note: all functions are calculated in log base 2, so return units of "bits". + +====== + +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 + +====== + +To compile the library for use in MATLAB/OCTAVE, execute CompileMIToolbox.m +from within MATLAB, or run 'make matlab' from a terminal. + +To compile the library for C/C++, run 'make' at a terminal. + +The C source files are licensed under the LGPL v3. The MATLAB wrappers and +demonstration feature selection algorithms are provided as is with no warranty +as examples of how to use the library in MATLAB. + +Update History +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 fault if a x by 0 empty matrix was passed in. Now prints an error message and returns gracefully +02/09/2010 - v1.01 - Updated CMIM.m in demonstration_algorithms, due to a bug where the last feature would not be selected first if it had the highest MI +07/07/2010 - v1.00 - Initial Release diff --git a/FEAST/FEAST/MIToolbox/RenyiEntropy.c b/FEAST/FEAST/MIToolbox/RenyiEntropy.c new file mode 100644 index 0000000..32c5ff9 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/RenyiEntropy.c @@ -0,0 +1,191 @@ +/******************************************************************************* +** RenyiEntropy.cpp +** Part of the mutual information toolbox +** +** Contains functions to calculate the Renyi alpha entropy of a single variable +** H_\alpha(X), the Renyi joint entropy of two variables H_\alpha(X,Y), and the +** conditional Renyi entropy H_\alpha(X|Y) +** +** Author: Adam Pocock +** Created 26/3/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#include "MIToolbox.h" +#include "ArrayOperations.h" +#include "CalculateProbability.h" +#include "Entropy.h" + +double calculateRenyiEntropy(double alpha, double *dataVector, int vectorLength) +{ + double entropy = 0.0; + double tempValue = 0.0; + int i; + ProbabilityState state = calculateProbability(dataVector,vectorLength); + + /*H_\alpha(X) = 1/(1-alpha) * log(2)(sum p(x)^alpha)*/ + for (i = 0; i < state.numStates; i++) + { + tempValue = state.probabilityVector[i]; + + if (tempValue > 0) + { + entropy += pow(tempValue,alpha); + /*printf("Entropy = %f, i = %d\n", entropy,i);*/ + } + } + + /*printf("Entropy = %f\n", entropy);*/ + + entropy = log(entropy); + + entropy /= log(2.0); + + entropy /= (1.0-alpha); + + /*printf("Entropy = %f\n", entropy);*/ + FREE_FUNC(state.probabilityVector); + state.probabilityVector = NULL; + + return entropy; +}/*calculateRenyiEntropy(double,double*,int)*/ + +double calculateJointRenyiEntropy(double alpha, double *firstVector, double *secondVector, int vectorLength) +{ + double jointEntropy = 0.0; + double tempValue = 0.0; + int i; + JointProbabilityState state = calculateJointProbability(firstVector,secondVector,vectorLength); + + /*H_\alpha(XY) = 1/(1-alpha) * log(2)(sum p(xy)^alpha)*/ + for (i = 0; i < state.numJointStates; i++) + { + tempValue = state.jointProbabilityVector[i]; + if (tempValue > 0) + { + jointEntropy += pow(tempValue,alpha); + } + } + + jointEntropy = log(jointEntropy); + + jointEntropy /= log(2.0); + + jointEntropy /= (1.0-alpha); + + FREE_FUNC(state.firstProbabilityVector); + state.firstProbabilityVector = NULL; + FREE_FUNC(state.secondProbabilityVector); + state.secondProbabilityVector = NULL; + FREE_FUNC(state.jointProbabilityVector); + state.jointProbabilityVector = NULL; + + return jointEntropy; +}/*calculateJointRenyiEntropy(double,double*,double*,int)*/ + +double calcCondRenyiEnt(double alpha, double *dataVector, double *conditionVector, int uniqueInCondVector, int vectorLength) +{ + /*uniqueInCondVector = is the number of unique values in the cond vector.*/ + + /*condEntropy = sum p(y) * sum p(x|y)^alpha(*/ + + /* + ** first generate the seperate variables + */ + + double *seperateVectors = (double *) CALLOC_FUNC(uniqueInCondVector*vectorLength,sizeof(double)); + int *seperateVectorCount = (int *) CALLOC_FUNC(uniqueInCondVector,sizeof(int)); + double seperateVectorProb = 0.0; + int i,j; + double entropy = 0.0; + double tempValue = 0.0; + int currentValue; + double tempEntropy; + ProbabilityState state; + + double **seperateVectors2D = (double **) CALLOC_FUNC(uniqueInCondVector,sizeof(double*)); + for(j=0; j < uniqueInCondVector; j++) + seperateVectors2D[j] = seperateVectors + (int)j*vectorLength; + + for (i = 0; i < vectorLength; i++) + { + currentValue = (int) (conditionVector[i] - 1.0); + /*printf("CurrentValue = %d\n",currentValue);*/ + seperateVectors2D[currentValue][seperateVectorCount[currentValue]] = dataVector[i]; + seperateVectorCount[currentValue]++; + } + + + + for (j = 0; j < uniqueInCondVector; j++) + { + tempEntropy = 0.0; + seperateVectorProb = ((double)seperateVectorCount[j]) / vectorLength; + state = calculateProbability(seperateVectors2D[j],seperateVectorCount[j]); + + /*H_\alpha(X) = 1/(1-alpha) * log(2)(sum p(x)^alpha)*/ + for (i = 0; i < state.numStates; i++) + { + tempValue = state.probabilityVector[i]; + + if (tempValue > 0) + { + tempEntropy += pow(tempValue,alpha); + /*printf("Entropy = %f, i = %d\n", entropy,i);*/ + } + } + + /*printf("Entropy = %f\n", entropy);*/ + + tempEntropy = log(tempEntropy); + + tempEntropy /= log(2.0); + + tempEntropy /= (1.0-alpha); + + entropy += tempEntropy; + + FREE_FUNC(state.probabilityVector); + } + + FREE_FUNC(seperateVectors2D); + seperateVectors2D = NULL; + + FREE_FUNC(seperateVectors); + FREE_FUNC(seperateVectorCount); + + seperateVectors = NULL; + seperateVectorCount = NULL; + + return entropy; +}/*calcCondRenyiEnt(double *,double *,int)*/ + +double calculateConditionalRenyiEntropy(double alpha, double *dataVector, double *conditionVector, int vectorLength) +{ + /*calls this: + **double calculateConditionalRenyiEntropy(double alpha, double *firstVector, double *condVector, int uniqueInCondVector, int vectorLength) + **after determining uniqueInCondVector + */ + int numUnique = numberOfUniqueValues(conditionVector, vectorLength); + + return calcCondRenyiEnt(alpha, dataVector, conditionVector, numUnique, vectorLength); +}/*calculateConditionalRenyiEntropy(double,double*,double*,int)*/ + diff --git a/FEAST/FEAST/MIToolbox/RenyiEntropy.h b/FEAST/FEAST/MIToolbox/RenyiEntropy.h new file mode 100644 index 0000000..296bc4b --- /dev/null +++ b/FEAST/FEAST/MIToolbox/RenyiEntropy.h @@ -0,0 +1,68 @@ +/******************************************************************************* +** RenyiEntropy.h +** Part of the mutual information toolbox +** +** Contains functions to calculate the Renyi alpha entropy of a single variable +** H_\alpha(X), the Renyi joint entropy of two variables H_\alpha(X,Y), and the +** conditional Renyi entropy H_\alpha(X|Y) +** +** Author: Adam Pocock +** Created 26/3/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#ifndef __Renyi_Entropy_H +#define __Renyi_Entropy_H + +#ifdef __cplusplus +extern "C" { +#endif + +/******************************************************************************* +** calculateRenyiEntropy returns the Renyi entropy in log base 2 of dataVector +** H_{\alpha}(X), for \alpha != 1 +** +** length(dataVector) == vectorLength otherwise there +** will be a segmentation fault +*******************************************************************************/ +double calculateRenyiEntropy(double alpha, double *dataVector, int vectorLength); + +/******************************************************************************* +** calculateJointRenyiEntropy returns the Renyi entropy in log base 2 of the +** joint variable of firstVector and secondVector H_{\alpha}(XY), +** for \alpha != 1 +** +** length(firstVector) == length(secondVector) == vectorLength otherwise there +** will be a segmentation fault +*******************************************************************************/ +double calculateJointRenyiEntropy(double alpha, double *firstVector, double *secondVector, int vectorLength); + +/* This function does not return a valid conditonal entropy as it has no +** meaning in Renyi's extension of entropy +double calculateConditionalRenyiEntropy(double alpha, double *dataVector, double *conditionVector, int vectorLength); +*/ + +#ifdef __cplusplus +} +#endif + +#endif + diff --git a/FEAST/FEAST/MIToolbox/RenyiMIToolbox.m b/FEAST/FEAST/MIToolbox/RenyiMIToolbox.m new file mode 100644 index 0000000..cd235e9 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/RenyiMIToolbox.m @@ -0,0 +1,48 @@ +function [varargout] = RenyiMIToolbox(functionName, alpha, varargin) +%function [varargout] = RenyiMIToolbox(functionName, alpha, varargin) +% +%Provides access to the functions in RenyiMIToolboxMex +% +%Expects column vectors, will not work with row vectors +% +%Function list +%"Entropy" = H_{\alpha}(X) = 1 +%"MI" = I_{\alpha}(X;Y) = 3 +% +%Arguments and returned values +%[entropy] = H_\alpha(X) = H(alpha,vector) +%[mi] = I_\alpha(X;Y) = I(alpha,vector,target) +% +%Internal RenyiMIToolbox function number +%Renyi Entropy = 1; +%Renyi MI = 3; + +if (alpha ~= 1) + if (strcmpi(functionName,'Entropy') || strcmpi(functionName,'h')) + %disp('Calculating Entropy'); + if (size(varargin{1},2)>1) + mergedVector = MIToolboxMex(3,varargin{1}); + else + mergedVector = varargin{1}; + end + [varargout{1}] = RenyiMIToolboxMex(1,alpha,mergedVector); + elseif (strcmpi(functionName,'MI')) + if (size(varargin{1},2)>1) + mergedFirst = MIToolboxMex(3,varargin{1}); + else + mergedFirst = varargin{1}; + end + if (size(varargin{2},2)>1) + mergedSecond = MIToolboxMex(3,varargin{2}); + else + mergedSecond = varargin{2}; + end + [varargout{1}] = RenyiMIToolboxMex(3,alpha,mergedFirst,mergedSecond); + else + varargout{1} = 0; + disp(['Unrecognised functionName ' functionName]); + end +else + disp('For alpha = 1 use functions in MIToolbox.m'); + disp('as those functions are the implementation of Shannon''s Information Theory'); +end diff --git a/FEAST/FEAST/MIToolbox/RenyiMIToolboxMex.c b/FEAST/FEAST/MIToolbox/RenyiMIToolboxMex.c new file mode 100644 index 0000000..03ab076 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/RenyiMIToolboxMex.c @@ -0,0 +1,197 @@ +/******************************************************************************* +** +** RenyiMIToolboxMex.cpp +** is the MATLAB entry point for the Renyi Entropy and MI MIToolbox functions +** when called from a MATLAB/OCTAVE script. +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ +#include "MIToolbox.h" +#include "RenyiEntropy.h" +#include "RenyiMutualInformation.h" + +/******************************************************************************* +**entry point for the mex call +**nlhs - number of outputs +**plhs - pointer to array of outputs +**nrhs - number of inputs +**prhs - pointer to array of inputs +*******************************************************************************/ +void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) +{ + /***************************************************************************** + ** this function takes a flag and 2 or 3 other arguments + ** the first is a scalar alpha value, and the remainder are + ** arrays. It returns a Renyi entropy or mutual information using the + ** alpha divergence. + *****************************************************************************/ + + int flag, numberOfSamples, checkSamples, numberOfFeatures, checkFeatures; + double alpha; + double *dataVector, *firstVector, *secondVector, *output; + + /*if (nlhs != 1) + { + printf("Incorrect number of output arguments\n"); + }//if not 1 output + */ + if (nrhs == 3) + { + /*printf("Must be H(X)\n");*/ + } + else if (nrhs == 4) + { + /*printf("Must be H(XY), I(X;Y)\n");*/ + } + else + { + printf("Incorrect number of arguments, format is RenyiMIToolbox(\"FLAG\",varargin)\n"); + } + + /* number to function map + ** 1 = H(X) + ** 2 = H(XY) + ** 3 = I(X;Y) + */ + + flag = *mxGetPr(prhs[0]); + + switch (flag) + { + case 1: + { + /* + **H_{\alpha}(X) + */ + alpha = mxGetScalar(prhs[1]); + numberOfSamples = mxGetM(prhs[2]); + numberOfFeatures = mxGetN(prhs[2]); + dataVector = (double *) mxGetPr(prhs[2]); + + plhs[0] = mxCreateDoubleMatrix(1,1,mxREAL); + output = (double *) mxGetPr(plhs[0]); + + if (numberOfFeatures == 1) + { + /*double calculateRenyiEntropy(double alpha, double *dataVector, long vectorLength);*/ + *output = calculateRenyiEntropy(alpha,dataVector,numberOfSamples); + } + else + { + printf("No columns in input\n"); + *output = -1.0; + } + break; + }/*case 1 - H_{\alpha}(X)*/ + case 2: + { + /* + **H_{\alpha}(XY) + */ + alpha = mxGetScalar(prhs[1]); + + numberOfSamples = mxGetM(prhs[2]); + checkSamples = mxGetM(prhs[3]); + + numberOfFeatures = mxGetN(prhs[2]); + checkFeatures = mxGetN(prhs[3]); + + firstVector = mxGetPr(prhs[2]); + secondVector = mxGetPr(prhs[3]); + + plhs[0] = mxCreateDoubleMatrix(1,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + if ((numberOfFeatures == 1) && (checkFeatures == 1)) + { + if ((numberOfSamples == 0) || (checkSamples == 0)) + { + *output = 0.0; + } + else if (numberOfSamples == checkSamples) + { + /*double calculateJointRenyiEntropy(double alpha, double *firstVector, double *secondVector, long vectorLength);*/ + *output = calculateJointRenyiEntropy(alpha,firstVector,secondVector,numberOfSamples); + } + else + { + printf("Vector lengths do not match, they must be the same length"); + *output = -1.0; + } + } + else + { + printf("No columns in input\n"); + *output = -1.0; + } + break; + }/*case 2 - H_{\alpha}(XY)*/ + case 3: + { + /* + **I_{\alpha}(X;Y) + */ + alpha = mxGetScalar(prhs[1]); + + numberOfSamples = mxGetM(prhs[2]); + checkSamples = mxGetM(prhs[3]); + + numberOfFeatures = mxGetN(prhs[2]); + checkFeatures = mxGetN(prhs[3]); + + firstVector = mxGetPr(prhs[2]); + secondVector = mxGetPr(prhs[3]); + + plhs[0] = mxCreateDoubleMatrix(1,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + if ((numberOfFeatures == 1) && (checkFeatures == 1)) + { + if ((numberOfSamples == 0) || (checkSamples == 0)) + { + *output = 0.0; + } + else if (numberOfSamples == checkSamples) + { + /*double calculateRenyiMIDivergence(double alpha, double *dataVector, double *targetVector, long vectorLength);*/ + *output = calculateRenyiMIDivergence(alpha,firstVector,secondVector,numberOfSamples); + } + else + { + printf("Vector lengths do not match, they must be the same length"); + *output = -1.0; + } + } + else + { + printf("No columns in input\n"); + *output = -1.0; + } + break; + }/*case 3 - I_{\alpha}(X;Y)*/ + default: + { + printf("Unrecognised flag\n"); + break; + }/*default*/ + }/*switch(flag)*/ + + return; +}/*mexFunction()*/ diff --git a/FEAST/FEAST/MIToolbox/RenyiMIToolboxMex.mexa64 b/FEAST/FEAST/MIToolbox/RenyiMIToolboxMex.mexa64 Binary files differnew file mode 100755 index 0000000..6bff1db --- /dev/null +++ b/FEAST/FEAST/MIToolbox/RenyiMIToolboxMex.mexa64 diff --git a/FEAST/FEAST/MIToolbox/RenyiMutualInformation.c b/FEAST/FEAST/MIToolbox/RenyiMutualInformation.c new file mode 100644 index 0000000..dc6fd51 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/RenyiMutualInformation.c @@ -0,0 +1,95 @@ +/******************************************************************************* +** RenyiMutualInformation.cpp +** Part of the mutual information toolbox +** +** Contains functions to calculate the Renyi mutual information of +** two variables X and Y, I_\alpha(X;Y), using the Renyi alpha divergence and +** the joint entropy difference +** +** Author: Adam Pocock +** Created 26/3/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#include "MIToolbox.h" +#include "CalculateProbability.h" +#include "RenyiEntropy.h" +#include "RenyiMutualInformation.h" + +double calculateRenyiMIDivergence(double alpha, double *dataVector, double *targetVector, int vectorLength) +{ + double mutualInformation = 0.0; + int firstIndex,secondIndex; + int i; + double jointTemp = 0.0; + double seperateTemp = 0.0; + double invAlpha = 1.0 - alpha; + JointProbabilityState state = calculateJointProbability(dataVector,targetVector,vectorLength); + + /* standard MI is D_KL(p(x,y)||p(x)p(y)) + ** which expands to + ** D_KL(p(x,y)||p(x)p(y)) = sum(p(x,y) * log(p(x,y)/(p(x)p(y)))) + ** + ** Renyi alpha divergence D_alpha(p(x,y)||p(x)p(y)) + ** expands to + ** D_alpha(p(x,y)||p(x)p(y)) = 1/(alpha-1) * log(sum((p(x,y)^alpha)*((p(x)p(y))^(1-alpha)))) + */ + + for (i = 0; i < state.numJointStates; i++) + { + firstIndex = i % state.numFirstStates; + secondIndex = i / state.numFirstStates; + + if ((state.jointProbabilityVector[i] > 0) && (state.firstProbabilityVector[firstIndex] > 0) && (state.secondProbabilityVector[secondIndex] > 0)) + { + jointTemp = pow(state.jointProbabilityVector[i],alpha); + seperateTemp = state.firstProbabilityVector[firstIndex] * state.secondProbabilityVector[secondIndex]; + seperateTemp = pow(seperateTemp,invAlpha); + mutualInformation += (jointTemp * seperateTemp); + } + } + + mutualInformation = log(mutualInformation); + mutualInformation /= log(2.0); + mutualInformation /= (alpha-1.0); + + FREE_FUNC(state.firstProbabilityVector); + state.firstProbabilityVector = NULL; + FREE_FUNC(state.secondProbabilityVector); + state.secondProbabilityVector = NULL; + FREE_FUNC(state.jointProbabilityVector); + state.jointProbabilityVector = NULL; + + return mutualInformation; +}/*calculateRenyiMIDivergence(double, double *, double *, int)*/ + +double calculateRenyiMIJoint(double alpha, double *dataVector, double *targetVector, int vectorLength) +{ + double hY = calculateRenyiEntropy(alpha, targetVector, vectorLength); + double hX = calculateRenyiEntropy(alpha, dataVector, vectorLength); + + double hXY = calculateJointRenyiEntropy(alpha, dataVector, targetVector, vectorLength); + + double answer = hX + hY - hXY; + + return answer; +}/*calculateRenyiMIJoint(double, double*, double*, int)*/ + diff --git a/FEAST/FEAST/MIToolbox/RenyiMutualInformation.h b/FEAST/FEAST/MIToolbox/RenyiMutualInformation.h new file mode 100644 index 0000000..07eff09 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/RenyiMutualInformation.h @@ -0,0 +1,60 @@ +/******************************************************************************* +** RenyiMutualInformation.h +** Part of the mutual information toolbox +** +** Contains functions to calculate the Renyi mutual information of +** two variables X and Y, I_\alpha(X;Y), using the Renyi alpha divergence and +** the joint entropy difference +** +** Author: Adam Pocock +** Created 26/3/2010 +** +** Copyright 2010 Adam Pocock, The University Of Manchester +** www.cs.manchester.ac.uk +** +** This file is part of MIToolbox. +** +** MIToolbox is free software: you can redistribute it and/or modify +** it under the terms of the GNU Lesser General Public License as published by +** the Free Software Foundation, either version 3 of the License, or +** (at your option) any later version. +** +** MIToolbox is distributed in the hope that it will be useful, +** but WITHOUT ANY WARRANTY; without even the implied warranty of +** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +** GNU Lesser General Public License for more details. +** +** You should have received a copy of the GNU Lesser General Public License +** along with MIToolbox. If not, see <http://www.gnu.org/licenses/>. +** +*******************************************************************************/ + +#ifndef __Renyi_MutualInformation_H +#define __Renyi_MutualInformation_H + +#ifdef __cplusplus +extern "C" { +#endif + +/******************************************************************************* +** calculateRenyiMIDivergence returns the log base 2 Renyi mutual information +** between dataVector and targetVector, I_{\alpha}(X;Y), for \alpha != 1 +** This uses Renyi's generalised alpha divergence as the difference measure +** instead of the KL-divergence as in Shannon's Mutual Information +** +** length(dataVector) == length(targetVector) == vectorLength otherwise there +** will be a segmentation fault +*******************************************************************************/ +double calculateRenyiMIDivergence(double alpha, double *dataVector, double *targetVector, int vectorLength); + +/* This function returns a different value to the alpha divergence mutual +** information, and thus is not a correct mutual information +double calculateRenyiMIJoint(double alpha, double *dataVector, double *targetVector, int vectorLength); +*/ + +#ifdef __cplusplus +} +#endif + +#endif + diff --git a/FEAST/FEAST/MIToolbox/cmi.m b/FEAST/FEAST/MIToolbox/cmi.m new file mode 100644 index 0000000..30e4bb0 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/cmi.m @@ -0,0 +1,31 @@ +function output = cmi(X,Y,Z) +%function output = cmi(X,Y,Z) +%X, Y & Z can be matrices which are converted into a joint variable +%before computation +% +%expects variables to be column-wise +% +%returns the mutual information between X and Y conditioned on Z, I(X;Y|Z) + +if nargin == 3 + if (size(X,2)>1) + mergedFirst = MIToolboxMex(3,X); + else + mergedFirst = X; + end + if (size(Y,2)>1) + mergedSecond = MIToolboxMex(3,Y); + else + mergedSecond = Y; + end + if (size(Z,2)>1) + mergedThird = MIToolboxMex(3,Z); + else + mergedThird = Z; + end + [output] = MIToolboxMex(8,mergedFirst,mergedSecond,mergedThird); +elseif nargin == 2 + output = mi(X,Y); +else + output = 0; +end diff --git a/FEAST/FEAST/MIToolbox/condh.m b/FEAST/FEAST/MIToolbox/condh.m new file mode 100644 index 0000000..9f966db --- /dev/null +++ b/FEAST/FEAST/MIToolbox/condh.m @@ -0,0 +1,26 @@ +function output = condh(X,Y) +%function output = condh(X,Y) +%X & Y can be matrices which are converted into a joint variable +%before computation +% +%expects variables to be column-wise +% +%returns the conditional entropy of X given Y, H(X|Y) + +if nargin == 2 + if (size(X,2)>1) + mergedFirst = MIToolboxMex(3,X); + else + mergedFirst = X; + end + if (size(Y,2)>1) + mergedSecond = MIToolboxMex(3,Y); + else + mergedSecond = Y; + end + [output] = MIToolboxMex(6,mergedFirst,mergedSecond); +elseif nargin == 1 + output = h(X); +else + output = 0; +end diff --git a/FEAST/FEAST/MIToolbox/demonstration_algorithms/CMIM.m b/FEAST/FEAST/MIToolbox/demonstration_algorithms/CMIM.m new file mode 100644 index 0000000..8ae1f7c --- /dev/null +++ b/FEAST/FEAST/MIToolbox/demonstration_algorithms/CMIM.m @@ -0,0 +1,49 @@ +function selectedFeatures = CMIM(k, featureMatrix, classColumn) +%function selectedFeatures = CMIM(k, featureMatrix, classColumn) +%Computes conditional mutual information maximisation algorithm from +%"Fast Binary Feature Selection with Conditional Mutual Information" +%by F. Fleuret (2004) + +%Computes the top k features from +%a dataset featureMatrix with n training examples and m features +%with the classes held in classColumn. + +noOfTraining = size(classColumn,1); +noOfFeatures = size(featureMatrix,2); + +partialScore = zeros(noOfFeatures,1); +m = zeros(noOfFeatures,1); +score = 0; +answerFeatures = zeros(k,1); +highestMI = 0; +highestMICounter = 0; + +for n = 1 : noOfFeatures + partialScore(n) = mi(featureMatrix(:,n),classColumn); + if partialScore(n) > highestMI + highestMI = partialScore(n); + highestMICounter = n; + end +end + +answerFeatures(1) = highestMICounter; + +for i = 2 : k + score = 0; + limitI = i - 1; + for n = 1 : noOfFeatures + while ((partialScore(n) >= score) && (m(n) < limitI)) + m(n) = m(n) + 1; + conditionalInfo = cmi(featureMatrix(:,n),classColumn,featureMatrix(:,answerFeatures(m(n)))); + if partialScore(n) > conditionalInfo + partialScore(n) = conditionalInfo; + end + end + if partialScore(n) >= score + score = partialScore(n); + answerFeatures(i) = n; + end + end +end + +selectedFeatures = answerFeatures; diff --git a/FEAST/FEAST/MIToolbox/demonstration_algorithms/CMIM_Mex.c b/FEAST/FEAST/MIToolbox/demonstration_algorithms/CMIM_Mex.c new file mode 100644 index 0000000..daacb34 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/demonstration_algorithms/CMIM_Mex.c @@ -0,0 +1,158 @@ +/******************************************************************************* +** Demonstration feature selection algorithm - MATLAB r2009a +** +** Initial Version - 13/06/2008 +** Updated - 07/07/2010 +** based on CMIM.m +** +** Conditional Mutual Information Maximisation +** in +** "Fast Binary Feature Selection using Conditional Mutual Information Maximisation +** F. Fleuret (2004) +** +** Author - Adam Pocock +** Demonstration code for MIToolbox +*******************************************************************************/ + +#include "mex.h" +#include "MutualInformation.h" + +void CMIMCalculation(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures) +{ + /*holds the class MI values + **the class MI doubles as the partial score from the CMIM paper + */ + double *classMI = (double *)mxCalloc(noOfFeatures,sizeof(double)); + /*in the CMIM paper, m = lastUsedFeature*/ + int *lastUsedFeature = (int *)mxCalloc(noOfFeatures,sizeof(int)); + + double score, conditionalInfo; + int iMinus, currentFeature; + + double maxMI = 0.0; + int maxMICounter = -1; + + int j,i; + + double **feature2D = (double**) mxCalloc(noOfFeatures,sizeof(double*)); + + for(j = 0; j < noOfFeatures; j++) + { + feature2D[j] = featureMatrix + (int)j*noOfSamples; + } + + for (i = 0; i < noOfFeatures;i++) + { + classMI[i] = calculateMutualInformation(feature2D[i], classColumn, noOfSamples); + + if (classMI[i] > maxMI) + { + maxMI = classMI[i]; + maxMICounter = i; + }/*if bigger than current maximum*/ + }/*for noOfFeatures - filling classMI*/ + + outputFeatures[0] = maxMICounter; + + /***************************************************************************** + ** We have populated the classMI array, and selected the highest + ** MI feature as the first output feature + ** Now we move into the CMIM algorithm + *****************************************************************************/ + + for (i = 1; i < k; i++) + { + score = 0.0; + iMinus = i-1; + + for (j = 0; j < noOfFeatures; j++) + { + while ((classMI[j] > score) && (lastUsedFeature[j] < i)) + { + /*double calculateConditionalMutualInformation(double *firstVector, double *targetVector, double *conditionVector, int vectorLength);*/ + currentFeature = (int) outputFeatures[lastUsedFeature[j]]; + conditionalInfo = calculateConditionalMutualInformation(feature2D[j],classColumn,feature2D[currentFeature],noOfSamples); + if (classMI[j] > conditionalInfo) + { + classMI[j] = conditionalInfo; + }/*reset classMI*/ + /*moved due to C indexing from 0 rather than 1*/ + lastUsedFeature[j] += 1; + }/*while partial score greater than score & not reached last feature*/ + if (classMI[j] > score) + { + score = classMI[j]; + outputFeatures[i] = j; + }/*if partial score still greater than score*/ + }/*for number of features*/ + }/*for the number of features to select*/ + + + for (i = 0; i < k; i++) + { + outputFeatures[i] += 1; /*C indexes from 0 not 1*/ + }/*for number of selected features*/ + +}/*CMIMCalculation*/ + +/*entry point for the mex call +**nlhs - number of outputs +**plhs - pointer to array of outputs +**nrhs - number of inputs +**prhs - pointer to array of inputs +*/ +void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) +{ + /************************************************************* + ** this function takes 3 arguments: + ** k = number of features to select, + ** featureMatrix[][] = matrix of features + ** classColumn[] = targets + ** the arguments should all be discrete integers. + ** and has one output: + ** selectedFeatures[] of size k + *************************************************************/ + + int k, numberOfFeatures, numberOfSamples, numberOfTargets; + double *featureMatrix, *targets, *output; + + + if (nlhs != 1) + { + printf("Incorrect number of output arguments"); + }/*if not 1 output*/ + if (nrhs != 3) + { + printf("Incorrect number of input arguments"); + }/*if not 3 inputs*/ + + /*get the number of features to select, cast out as it is a double*/ + k = (int) mxGetScalar(prhs[0]); + + numberOfFeatures = mxGetN(prhs[1]); + numberOfSamples = mxGetM(prhs[1]); + + numberOfTargets = mxGetM(prhs[2]); + + if (numberOfTargets != numberOfSamples) + { + printf("Number of targets must match number of samples\n"); + printf("Number of targets = %d, Number of Samples = %d, Number of Features = %d\n",numberOfTargets,numberOfSamples,numberOfFeatures); + + plhs[0] = mxCreateDoubleMatrix(0,0,mxREAL); + }/*if size mismatch*/ + else + { + + featureMatrix = mxGetPr(prhs[1]); + targets = mxGetPr(prhs[2]); + + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /*void CMIMCalculation(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures)*/ + CMIMCalculation(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output); + } + + return; +}/*mexFunction()*/ diff --git a/FEAST/FEAST/MIToolbox/demonstration_algorithms/DISR.m b/FEAST/FEAST/MIToolbox/demonstration_algorithms/DISR.m new file mode 100644 index 0000000..c8f5669 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/demonstration_algorithms/DISR.m @@ -0,0 +1,73 @@ +function selectedFeatures = DISR(k, featureMatrix, classColumn)
+%function selectedFeatures = DISR(k, featureMatrix, classColumn)
+%
+%Computers optimal features according to DISR algorithm from
+%On the Use of variable "complementarity for feature selection"
+%by P Meyer, G Bontempi (2006)
+%
+%Computes the top k features from
+%a dataset featureMatrix with n training examples and m features
+%with the classes held in classColumn.
+%
+%DISR - arg(Xi) max(sum(Xj mem XS)(SimRel(Xij,Y)))
+%where SimRel = MI(Xij,Y) / H(Xij,Y)
+
+totalFeatures = size(featureMatrix,2);
+classMI = zeros(totalFeatures,1);
+unselectedFeatures = ones(totalFeatures,1);
+score = 0;
+currentScore = 0;
+innerScore = 0;
+iMinus = 0;
+answerFeatures = zeros(k,1);
+highestMI = 0;
+highestMICounter = 0;
+currentHighestFeature = 0;
+
+%create a matrix to hold the SRs of a feature pair.
+%initialised to -1 as you can't get a negative SR.
+featureSRMatrix = -(ones(k,totalFeatures));
+
+for n = 1 : totalFeatures
+ classMI(n) = mi(featureMatrix(:,n),classColumn);
+ if classMI(n) > highestMI
+ highestMI = classMI(n);
+ highestMICounter = n;
+ end
+end
+
+answerFeatures(1) = highestMICounter;
+unselectedFeatures(highestMICounter) = 0;
+
+for i = 2 : k
+ score = 0;
+ currentHighestFeature = 0;
+ iMinus = i-1;
+ for j = 1 : totalFeatures
+ if unselectedFeatures(j) == 1
+ %DISR - arg(Xi) max(sum(Xj mem XS)(SimRel(Xij,Y)))
+ %where SimRel = MI(Xij,Y) / H(Xij,Y)
+ currentScore = 0;
+ for m = 1 : iMinus
+ if featureSRMatrix(m,j) == -1
+ unionedFeatures = joint([featureMatrix(:,answerFeatures(m)),featureMatrix(:,j)]);
+ tempUnionMI = mi(unionedFeatures,classColumn);
+ tempTripEntropy = h([unionedFeatures,classColumn]);
+ featureSRMatrix(m,j) = tempUnionMI/tempTripEntropy;
+ end
+
+ currentScore = currentScore + featureSRMatrix(m,j);
+ end
+ if (currentScore > score)
+ score = currentScore;
+ currentHighestFeature = j;
+ end
+ end
+ end
+ %now highest feature is selected in currentHighestFeature
+ %store it
+ unselectedFeatures(currentHighestFeature) = 0;
+ answerFeatures(i) = currentHighestFeature;
+end
+
+selectedFeatures = answerFeatures;
diff --git a/FEAST/FEAST/MIToolbox/demonstration_algorithms/DISR_Mex.c b/FEAST/FEAST/MIToolbox/demonstration_algorithms/DISR_Mex.c new file mode 100644 index 0000000..617ca81 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/demonstration_algorithms/DISR_Mex.c @@ -0,0 +1,199 @@ +/******************************************************************************* +** Demonstration feature selection algorithm - MATLAB r2009a +** +** Initial Version - 13/06/2008 +** Updated - 07/07/2010 +** based on DISR.m +** +** Double Input Symmetrical Relevance +** in +** "On the Use of Variable Complementarity for Feature Selection in Cancer Classification" +** P. Meyer and G. Bontempi (2006) +** +** Author - Adam Pocock +** Demonstration code for MIToolbox +*******************************************************************************/ + +#include "mex.h" +#include "MutualInformation.h" +#include "Entropy.h" +#include "ArrayOperations.h" + +void DISRCalculation(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures) +{ + /*holds the class MI values*/ + double *classMI = (double *)mxCalloc(noOfFeatures,sizeof(double)); + + char *selectedFeatures = (char *)mxCalloc(noOfFeatures,sizeof(char)); + + /*holds the intra feature MI values*/ + int sizeOfMatrix = k*noOfFeatures; + double *featureMIMatrix = (double *)mxCalloc(sizeOfMatrix,sizeof(double)); + + double maxMI = 0.0; + int maxMICounter = -1; + + double **feature2D = (double**) mxCalloc(noOfFeatures,sizeof(double*)); + + double score, currentScore, totalFeatureMI; + int currentHighestFeature; + + double *mergedVector = (double *) mxCalloc(noOfSamples,sizeof(double)); + + int arrayPosition; + double mi, tripEntropy; + + int i,j,x; + + for(j = 0; j < noOfFeatures; j++) + { + feature2D[j] = featureMatrix + (int)j*noOfSamples; + } + + for (i = 0; i < sizeOfMatrix;i++) + { + featureMIMatrix[i] = -1; + }/*for featureMIMatrix - blank to -1*/ + + + for (i = 0; i < noOfFeatures;i++) + { + /*calculate mutual info + **double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength); + */ + classMI[i] = calculateMutualInformation(feature2D[i], classColumn, noOfSamples); + + if (classMI[i] > maxMI) + { + maxMI = classMI[i]; + maxMICounter = i; + }/*if bigger than current maximum*/ + }/*for noOfFeatures - filling classMI*/ + + selectedFeatures[maxMICounter] = 1; + outputFeatures[0] = maxMICounter; + + /***************************************************************************** + ** We have populated the classMI array, and selected the highest + ** MI feature as the first output feature + ** Now we move into the DISR algorithm + *****************************************************************************/ + + for (i = 1; i < k; i++) + { + score = 0.0; + currentHighestFeature = 0; + currentScore = 0.0; + totalFeatureMI = 0.0; + + for (j = 0; j < noOfFeatures; j++) + { + /*if we haven't selected j*/ + if (selectedFeatures[j] == 0) + { + currentScore = 0.0; + totalFeatureMI = 0.0; + + for (x = 0; x < i; x++) + { + arrayPosition = x*noOfFeatures + j; + if (featureMIMatrix[arrayPosition] == -1) + { + /* + **double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength); + **double calculateJointEntropy(double *firstVector, double *secondVector, int vectorLength); + */ + + mergeArrays(feature2D[(int) outputFeatures[x]], feature2D[j],mergedVector,noOfSamples); + mi = calculateMutualInformation(mergedVector, classColumn, noOfSamples); + tripEntropy = calculateJointEntropy(mergedVector, classColumn, noOfSamples); + + featureMIMatrix[arrayPosition] = mi / tripEntropy; + }/*if not already known*/ + currentScore += featureMIMatrix[arrayPosition]; + }/*for the number of already selected features*/ + + if (currentScore > score) + { + score = currentScore; + currentHighestFeature = j; + } + }/*if j is unselected*/ + }/*for number of features*/ + + selectedFeatures[currentHighestFeature] = 1; + outputFeatures[i] = currentHighestFeature; + + }/*for the number of features to select*/ + + mxFree(mergedVector); + mergedVector = NULL; + + for (i = 0; i < k; i++) + { + outputFeatures[i] += 1; /*C indexes from 0 not 1*/ + }/*for number of selected features*/ + +}/*DISRCalculation(double[][],double[])*/ + +/*entry point for the mex call +**nlhs - number of outputs +**plhs - pointer to array of outputs +**nrhs - number of inputs +**prhs - pointer to array of inputs +*/ +void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) +{ + /************************************************************* + ** this function takes 3 arguments: + ** k = number of features to select, + ** featureMatrix[][] = matrix of features + ** classColumn[] = targets + ** the arguments should all be discrete integers. + ** and has one output: + ** selectedFeatures[] of size k + *************************************************************/ + + int k, numberOfFeatures, numberOfSamples, numberOfTargets; + double *featureMatrix, *targets, *output; + + + if (nlhs != 1) + { + printf("Incorrect number of output arguments"); + }/*if not 1 output*/ + if (nrhs != 3) + { + printf("Incorrect number of input arguments"); + }/*if not 3 inputs*/ + + /*get the number of features to select, cast out as it is a double*/ + k = (int) mxGetScalar(prhs[0]); + + numberOfFeatures = mxGetN(prhs[1]); + numberOfSamples = mxGetM(prhs[1]); + + numberOfTargets = mxGetM(prhs[2]); + + if (numberOfTargets != numberOfSamples) + { + printf("Number of targets must match number of samples\n"); + printf("Number of targets = %d, Number of Samples = %d, Number of Features = %d\n",numberOfTargets,numberOfSamples,numberOfFeatures); + + plhs[0] = mxCreateDoubleMatrix(0,0,mxREAL); + }/*if size mismatch*/ + else + { + + featureMatrix = mxGetPr(prhs[1]); + targets = mxGetPr(prhs[2]); + + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /*void DISRCalculation(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures)*/ + DISRCalculation(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output); + } + + return; +}/*mexFunction()*/ diff --git a/FEAST/FEAST/MIToolbox/demonstration_algorithms/IAMB.m b/FEAST/FEAST/MIToolbox/demonstration_algorithms/IAMB.m new file mode 100644 index 0000000..1011260 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/demonstration_algorithms/IAMB.m @@ -0,0 +1,56 @@ +function [cmb association] = IAMB( data, targetindex, THRESHOLD) +%function [cmb association] = IAMB( data, targetindex, THRESHOLD) +% +%Performs the IAMB algorithm of Tsmardinos et al. (2003) +%from "Towards principled feature selection: Relevancy, filters and wrappers" + +if (nargin == 2) + THRESHOLD = 0.02; +end + +numf = size(data,2); +targets = data(:,targetindex); +data(:,targetindex) = -10; + +cmb = []; + +finished = false; +while ~finished + for n = 1:numf + cmbVector = joint(data(:,cmb)); + if isempty(cmb) + association(n) = mi( data(:,n), targets ); + end + + if ismember(n,cmb) + association(n) = -10; %arbtirary large negative constant + else + association(n) = cmi( data(:,n), targets, cmbVector); + end + end + + [maxval maxidx] = max(association); + if maxval < THRESHOLD + finished = true; + else + cmb = [ cmb maxidx ]; + end +end + +finished = false; +while ~finished && ~isempty(cmb) + association = []; + for n = 1:length(cmb) + cmbwithoutn = cmb; + cmbwithoutn(n)=[]; + association(n) = cmi( data(:,cmb(n)), targets, data(:,cmbwithoutn) ); + end + + [minval minidx] = min(association); + if minval > THRESHOLD + finished = true; + else + cmb(minidx) = []; + end +end + diff --git a/FEAST/FEAST/MIToolbox/demonstration_algorithms/compile_demos.m b/FEAST/FEAST/MIToolbox/demonstration_algorithms/compile_demos.m new file mode 100644 index 0000000..65464b3 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/demonstration_algorithms/compile_demos.m @@ -0,0 +1,3 @@ +mex -I.. CMIM_Mex.c ../MutualInformation.c ../Entropy.c ../CalculateProbability.c ../ArrayOperations.c +mex -I.. DISR_Mex.c ../MutualInformation.c ../Entropy.c ../CalculateProbability.c ../ArrayOperations.c +mex -I.. mRMR_D_Mex.c ../MutualInformation.c ../Entropy.c ../CalculateProbability.c ../ArrayOperations.c
\ No newline at end of file diff --git a/FEAST/FEAST/MIToolbox/demonstration_algorithms/mRMR_D.m b/FEAST/FEAST/MIToolbox/demonstration_algorithms/mRMR_D.m new file mode 100644 index 0000000..50b14bc --- /dev/null +++ b/FEAST/FEAST/MIToolbox/demonstration_algorithms/mRMR_D.m @@ -0,0 +1,69 @@ +function selectedFeatures = mRMR_D(k, featureMatrix, classColumn) +%function selectedFeatures = mRMR_D(k, featureMatrix, classColumn) +% +%Selects optimal features according to the mRMR-D algorithm from +%"Feature Selection Based on Mutual Information: Criteria of Max-Dependency, Max-Relevance, and Min-Redundancy" +%by H. Peng et al. (2005) +% +%Calculates the top k features +%a dataset featureMatrix with n training examples and m features +%with the classes held in classColumn (an n x 1 vector) + +noOfTraining = size(classColumn,1); +noOfFeatures = size(featureMatrix,2); +unselectedFeatures = ones(noOfFeatures,1); + +classMI = zeros(noOfFeatures,1); +answerFeatures = zeros(k,1); +highestMI = 0; +highestMICounter = 0; +currentHighestFeature = 0; + +featureMIMatrix = -(ones(k,noOfFeatures)); + +%setup the mi against the class +for n = 1 : noOfFeatures + classMI(n) = mi(featureMatrix(:,n),classColumn); + if classMI(n) > highestMI + highestMI = classMI(n); + highestMICounter = n; + end +end + +answerFeatures(1) = highestMICounter; +unselectedFeatures(highestMICounter) = 0; + +%iterate over the number of features to select +for i = 2:k + score = -100; + currentHighestFeature = 0; + iMinus = i-1; + for j = 1 : noOfFeatures + if unselectedFeatures(j) == 1 + currentMIScore = 0; + for m = 1 : iMinus + if featureMIMatrix(m,j) == -1 + featureMIMatrix(m,j) = mi(featureMatrix(:,j),featureMatrix(:,answerFeatures(m))); + end + currentMIScore = currentMIScore + featureMIMatrix(m,j); + end + currentScore = classMI(j) - (currentMIScore/iMinus); + + if (currentScore > score) + score = currentScore; + currentHighestFeature = j; + end + end + end + + if score < 0 + disp(['at selection ' int2str(j) ' mRMRD is negative with value ' num2str(score)]); + end + + %now highest feature is selected in currentHighestFeature + %store it + unselectedFeatures(currentHighestFeature) = 0; + answerFeatures(i) = currentHighestFeature; +end + +selectedFeatures = answerFeatures; diff --git a/FEAST/FEAST/MIToolbox/demonstration_algorithms/mRMR_D_Mex.c b/FEAST/FEAST/MIToolbox/demonstration_algorithms/mRMR_D_Mex.c new file mode 100644 index 0000000..8d7b074 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/demonstration_algorithms/mRMR_D_Mex.c @@ -0,0 +1,184 @@ +/******************************************************************************* +** Demonstration feature selection algorithm - MATLAB r2009a +** +** Initial Version - 13/06/2008 +** Updated - 07/07/2010 +** based on mRMR_D.m +** +** Minimum Relevance Maximum Redundancy +** in +** "Feature Selection Based on Mutual Information: Criteria of Max-Dependency, Max-Relevance, and Min-Redundancy" +** H. Peng et al. (2005) +** +** Author - Adam Pocock +** Demonstration code for MIToolbox +*******************************************************************************/ + +#include "mex.h" +#include "MutualInformation.h" + +void mRMRCalculation(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures) +{ + double **feature2D = (double**) mxCalloc(noOfFeatures,sizeof(double*)); + /*holds the class MI values*/ + double *classMI = (double *)mxCalloc(noOfFeatures,sizeof(double)); + int *selectedFeatures = (int *)mxCalloc(noOfFeatures,sizeof(int)); + /*holds the intra feature MI values*/ + int sizeOfMatrix = k*noOfFeatures; + double *featureMIMatrix = (double *)mxCalloc(sizeOfMatrix,sizeof(double)); + + double maxMI = 0.0; + int maxMICounter = -1; + + /*init variables*/ + + double score, currentScore, totalFeatureMI; + int currentHighestFeature; + + int arrayPosition, i, j, x; + + for(j = 0; j < noOfFeatures; j++) + { + feature2D[j] = featureMatrix + (int)j*noOfSamples; + } + + for (i = 0; i < sizeOfMatrix;i++) + { + featureMIMatrix[i] = -1; + }/*for featureMIMatrix - blank to -1*/ + + + for (i = 0; i < noOfFeatures;i++) + { + classMI[i] = calculateMutualInformation(feature2D[i], classColumn, noOfSamples); + if (classMI[i] > maxMI) + { + maxMI = classMI[i]; + maxMICounter = i; + }/*if bigger than current maximum*/ + }/*for noOfFeatures - filling classMI*/ + + selectedFeatures[maxMICounter] = 1; + outputFeatures[0] = maxMICounter; + + /************* + ** Now we have populated the classMI array, and selected the highest + ** MI feature as the first output feature + ** Now we move into the mRMR-D algorithm + *************/ + + for (i = 1; i < k; i++) + { + /*to ensure it selects some features + **if this is zero then it will not pick features where the redundancy is greater than the + **relevance + */ + score = -1000.0; + currentHighestFeature = 0; + currentScore = 0.0; + totalFeatureMI = 0.0; + + for (j = 0; j < noOfFeatures; j++) + { + /*if we haven't selected j*/ + if (selectedFeatures[j] == 0) + { + currentScore = classMI[j]; + totalFeatureMI = 0.0; + + for (x = 0; x < i; x++) + { + arrayPosition = x*noOfFeatures + j; + if (featureMIMatrix[arrayPosition] == -1) + { + /*work out intra MI*/ + + /*double calculateMutualInformation(double *firstVector, double *secondVector, int vectorLength);*/ + featureMIMatrix[arrayPosition] = calculateMutualInformation(feature2D[(int) outputFeatures[x]], feature2D[j], noOfSamples); + } + + totalFeatureMI += featureMIMatrix[arrayPosition]; + }/*for the number of already selected features*/ + + currentScore -= (totalFeatureMI/i); + if (currentScore > score) + { + score = currentScore; + currentHighestFeature = j; + } + }/*if j is unselected*/ + }/*for number of features*/ + + selectedFeatures[currentHighestFeature] = 1; + outputFeatures[i] = currentHighestFeature; + + }/*for the number of features to select*/ + + for (i = 0; i < k; i++) + { + outputFeatures[i] += 1; /*C indexes from 0 not 1*/ + }/*for number of selected features*/ + +}/*mRMRCalculation(double[][],double[])*/ + +/*entry point for the mex call +**nlhs - number of outputs +**plhs - pointer to array of outputs +**nrhs - number of inputs +**prhs - pointer to array of inputs +*/ +void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) +{ + /************************************************************* + ** this function takes 3 arguments: + ** k = number of features to select, + ** featureMatrix[][] = matrix of features + ** classColumn[] = targets + ** the arguments should all be discrete integers. + ** and has one output: + ** selectedFeatures[] of size k + *************************************************************/ + + int k, numberOfFeatures, numberOfSamples, numberOfTargets; + double *featureMatrix, *targets, *output; + + + if (nlhs != 1) + { + printf("Incorrect number of output arguments"); + }/*if not 1 output*/ + if (nrhs != 3) + { + printf("Incorrect number of input arguments"); + }/*if not 3 inputs*/ + + /*get the number of features to select, cast out as it is a double*/ + k = (int) mxGetScalar(prhs[0]); + + numberOfFeatures = mxGetN(prhs[1]); + numberOfSamples = mxGetM(prhs[1]); + + numberOfTargets = mxGetM(prhs[2]); + + if (numberOfTargets != numberOfSamples) + { + printf("Number of targets must match number of samples\n"); + printf("Number of targets = %d, Number of Samples = %d, Number of Features = %d\n",numberOfTargets,numberOfSamples,numberOfFeatures); + + plhs[0] = mxCreateDoubleMatrix(0,0,mxREAL); + }/*if size mismatch*/ + else + { + + featureMatrix = mxGetPr(prhs[1]); + targets = mxGetPr(prhs[2]); + + plhs[0] = mxCreateDoubleMatrix(k,1,mxREAL); + output = (double *)mxGetPr(plhs[0]); + + /*void mRMRCalculation(int k, int noOfSamples, int noOfFeatures,double *featureMatrix, double *classColumn, double *outputFeatures)*/ + mRMRCalculation(k,numberOfSamples,numberOfFeatures,featureMatrix,targets,output); + } + + return; +}/*mexFunction()*/ diff --git a/FEAST/FEAST/MIToolbox/h.m b/FEAST/FEAST/MIToolbox/h.m new file mode 100644 index 0000000..8fd8999 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/h.m @@ -0,0 +1,13 @@ +function output = h(X) +%function output = h(X) +%X can be a matrix which is converted into a joint variable before calculation +%expects variables to be column-wise +% +%returns the entropy of X, H(X) + +if (size(X,2)>1) + mergedVector = MIToolboxMex(3,X); +else + mergedVector = X; +end +[output] = MIToolboxMex(4,mergedVector); diff --git a/FEAST/FEAST/MIToolbox/joint.m b/FEAST/FEAST/MIToolbox/joint.m new file mode 100644 index 0000000..f40ff25 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/joint.m @@ -0,0 +1,16 @@ +function output = joint(X,arities) +%function output = joint(X,arities) +%returns the joint random variable of the matrix X +%assuming the variables are in columns +% +%if passed a vector of the arities then it produces a correct +%joint variable, otherwise it may not include all states +% +%if the joint variable is only compared with variables using the same samples, +%then arity information is not required + +if (nargin == 2) + [output] = MIToolboxMex(3,X,arities); +else + [output] = MIToolboxMex(3,X); +end diff --git a/FEAST/FEAST/MIToolbox/mi.m b/FEAST/FEAST/MIToolbox/mi.m new file mode 100644 index 0000000..2fd8766 --- /dev/null +++ b/FEAST/FEAST/MIToolbox/mi.m @@ -0,0 +1,20 @@ +function output = mi(X,Y) +%function output = mi(X,Y) +%X & Y can be matrices which are converted into a joint variable +%before computation +% +%expects variables to be column-wise +% +%returns the mutual information between X and Y, I(X;Y) + +if (size(X,2)>1) + mergedFirst = MIToolboxMex(3,X); +else + mergedFirst = X; +end +if (size(Y,2)>1) + mergedSecond = MIToolboxMex(3,Y); +else + mergedSecond = Y; +end +[output] = MIToolboxMex(7,mergedFirst,mergedSecond); |