rftreenode.cpp

   1 //
   2 //  rftreenode.cpp
   3 //  Mothur
   4 //
   5 //  Created by Sarah Westcott on 10/2/12.
   6 //  Copyright (c) 2012 Schloss Lab. All rights reserved.
   7 //
   8
   9 #include "rftreenode.hpp"
  10
  11 /***********************************************************************/
  12 RFTreeNode::RFTreeNode(vector< vector<int> > bootstrappedTrainingSamples,
  13            vector<int> globalDiscardedFeatureIndices,
  14            int numFeatures,
  15            int numSamples,
  16            int numOutputClasses,
  17            int generation)
  18
  19 : bootstrappedTrainingSamples(bootstrappedTrainingSamples),
  20 globalDiscardedFeatureIndices(globalDiscardedFeatureIndices),
  21 numFeatures(numFeatures),
  22 numSamples(numSamples),
  23 numOutputClasses(numOutputClasses),
  24 generation(generation),
  25 isLeaf(false),
  26 outputClass(-1),
  27 splitFeatureIndex(-1),
  28 splitFeatureValue(-1),
  29 splitFeatureEntropy(-1.0),
  30 ownEntropy(-1.0),
  31 bootstrappedFeatureVectors(numFeatures, vector<int>(numSamples, 0)),
  32 bootstrappedOutputVector(numSamples, 0),
  33 leftChildNode(NULL),
  34 rightChildNode(NULL),
  35 parentNode(NULL) {
  36     m = MothurOut::getInstance();
  37
  38     for (int i = 0; i < numSamples; i++) {    // just doing a simple transpose of the matrix
  39         if (m->control_pressed) { break; }
  40         for (int j = 0; j < numFeatures; j++) { bootstrappedFeatureVectors[j][i] = bootstrappedTrainingSamples[i][j]; }
  41     }
  42
  43     for (int i = 0; i < numSamples; i++) { if (m->control_pressed) { break; } bootstrappedOutputVector[i] = bootstrappedTrainingSamples[i][numFeatures]; }
  44
  45     createLocalDiscardedFeatureList();
  46     updateNodeEntropy();
  47 }
  48 /***********************************************************************/
  49 int RFTreeNode::createLocalDiscardedFeatureList(){
  50     try {
  51
  52         for (int i = 0; i < numFeatures; i++) {
  53             if (m->control_pressed) { return 0; }
  54             vector<int>::iterator it = find(globalDiscardedFeatureIndices.begin(), globalDiscardedFeatureIndices.end(), i);
  55             if (it == globalDiscardedFeatureIndices.end()){                           // NOT FOUND
  56                 double standardDeviation = m->getStandardDeviation(bootstrappedFeatureVectors[i]);
  57                 if (standardDeviation <= 0){ localDiscardedFeatureIndices.push_back(i); }
  58             }
  59         }
  60
  61         return 0;
  62     }
  63     catch(exception& e) {
  64         m->errorOut(e, "RFTreeNode", "createLocalDiscardedFeatureList");
  65         exit(1);
  66     }
  67 }
  68 /***********************************************************************/
  69 int RFTreeNode::updateNodeEntropy() {
  70     try {
  71
  72         vector<int> classCounts(numOutputClasses, 0);
  73         for (int i = 0; i < bootstrappedOutputVector.size(); i++) { classCounts[bootstrappedOutputVector[i]]++; }
  74         int totalClassCounts = accumulate(classCounts.begin(), classCounts.end(), 0);
  75         double nodeEntropy = 0.0;
  76         for (int i = 0; i < classCounts.size(); i++) {
  77             if (m->control_pressed) { return 0; }
  78             if (classCounts[i] == 0) continue;
  79             double probability = (double)classCounts[i] / (double)totalClassCounts;
  80             nodeEntropy += -(probability * log2(probability));
  81         }
  82         ownEntropy = nodeEntropy;
  83
  84         return 0;
  85     }
  86     catch(exception& e) {
  87         m->errorOut(e, "RFTreeNode", "updateNodeEntropy");
  88         exit(1);
  89     }
  90 }
  91
  92 /***********************************************************************/