#include "amma/Num/NumClassifyBayes.hh"
#include "amma/Num/NumClassifyFuzzyCmeans.hh"
#include "amma/Num/NumPDFNormal.hh"
#include "amma/Num/NumPerfHoldOut.hh"
#include "amma/Num/NumPerfCrossVal.hh"
#include "amma/Num/NumPerfLeaveOutOne.hh"
#include "amma/Num/NumPerfRotation.hh"
#include "amma/Num/NumErrorCount.hh"
#include "amma/Num/NumErrorWeight.hh"
#include "amma/Num/NumClustMinVariance.hh"
#include "amma/Num/NumClustScatterDet.hh"
#include "amma/Num/NumMagnitudeEuclidean.hh"
#include "amma/Vector2d.hh"
#include <fstream.h>

main ()
{
  // Setup training data
  NumVLDataSetC trainL;
  trainL.Append (Vector2dC(1,1),0);
  trainL.Append (Vector2dC(1,-1),0);
  trainL.Append (Vector2dC(2,0),0);
  trainL.Append (Vector2dC(1.5,0.5),0);
  trainL.Append (Vector2dC(-1,1),1);
  trainL.Append (Vector2dC(-1,-1),1);
  trainL.Append (Vector2dC(-2,0),1);
  trainL.Append (Vector2dC(-1.5,-0.5),1);
  // Setup classifier stuff
  NumPDFNormalC pdf;
  VectorC prior (2); prior.Fill (0.5);
  NumClassifyBayesC class1 (pdf,prior);
  NumClassifyFuzzyCmeansC class2 (2,NumMagnitudeEuclideanC());
  // Setup error weighting stuff
  NumErrorCountC errorCount;
  // Setup performance evaluators
  SArray1dC<NumPerformanceC> eval (4);
  eval[0] = NumPerfHoldOutC (0.333);
  eval[1] = NumPerfCrossValC (2);
  eval[2] = NumPerfLeaveOutOneC ();
  eval[3] = NumPerfRotationC (2);
  // Setup clustering criteria
  SArray1dC<NumClustCriterionC> crit (2);
  crit[0] = NumClustMinVarianceC (NumMagnitudeEuclideanC());
  crit[1] = NumClustScatterDetC ();
  // Test classifier
  FOR_SARRAY1 (eval,i) {
    cout << eval[i].GetInfo () << "\n";
    cout << "PMC = " << eval[i].PMC (trainL,class1,errorCount) << "\n";
  }
  // Test clustering
  FOR_SARRAY1 (crit,i2) {
    cout << crit[i2].GetInfo () << endl;
    for (UIntT n = 1; n < 8; n++) {
      class2.Initialise (n,trainL.InputSet());
      class2.Cluster (trainL.InputSet());
      cout << "Clustering quality (" << n << ") = " << crit[i2].Criterion (trainL.InputSet(),class2) << endl;
    }
  }
}