TMVAMulticlass_MLP.class.C

Go to the documentation of this file.

// Class: ReadMLP
// Automatically generated by MethodBase::MakeClass
//

/* configuration options =====================================================

#GEN -*-*-*-*-*-*-*-*-*-*-*- general info -*-*-*-*-*-*-*-*-*-*-*-

Method         : MLP::MLP
TMVA Release   : 4.2.1         [262657]
ROOT Release   : 6.12/07       [396295]
Creator        : sftnight
Date           : Sat Sep 29 23:27:12 2018
Host           : Linux ec-ubuntu-14-04-x86-64-2 3.13.0-157-generic #207-Ubuntu SMP Mon Aug 20 16:44:59 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux
Dir            : /mnt/build/workspace/root-makedoc-v612/rootspi/rdoc/src/v6-12-00-patches/documentation/doxygen
Training events: 4000
Analysis type  : [Classification]


#OPT -*-*-*-*-*-*-*-*-*-*-*-*- options -*-*-*-*-*-*-*-*-*-*-*-*-

# Set by User:
NCycles: "1000" [Number of training cycles]
HiddenLayers: "N+5,5" [Specification of hidden layer architecture]
NeuronType: "tanh" [Neuron activation function type]
EstimatorType: "MSE" [MSE (Mean Square Estimator) for Gaussian Likelihood or CE(Cross-Entropy) for Bernoulli Likelihood]
V: "False" [Verbose output (short form of "VerbosityLevel" below - overrides the latter one)]
H: "False" [Print method-specific help message]
TestRate: "5" [Test for overtraining performed at each #th epochs]
# Default:
RandomSeed: "1" [Random seed for initial synapse weights (0 means unique seed for each run; default value '1')]
NeuronInputType: "sum" [Neuron input function type]
VerbosityLevel: "Default" [Verbosity level]
VarTransform: "None" [List of variable transformations performed before training, e.g., "D_Background,P_Signal,G,N_AllClasses" for: "Decorrelation, PCA-transformation, Gaussianisation, Normalisation, each for the given class of events ('AllClasses' denotes all events of all classes, if no class indication is given, 'All' is assumed)"]
CreateMVAPdfs: "False" [Create PDFs for classifier outputs (signal and background)]
IgnoreNegWeightsInTraining: "False" [Events with negative weights are ignored in the training (but are included for testing and performance evaluation)]
TrainingMethod: "BP" [Train with Back-Propagation (BP), BFGS Algorithm (BFGS), or Genetic Algorithm (GA - slower and worse)]
LearningRate: "2.000000e-02" [ANN learning rate parameter]
DecayRate: "1.000000e-02" [Decay rate for learning parameter]
EpochMonitoring: "False" [Provide epoch-wise monitoring plots according to TestRate (caution: causes big ROOT output file!)]
Sampling: "1.000000e+00" [Only 'Sampling' (randomly selected) events are trained each epoch]
SamplingEpoch: "1.000000e+00" [Sampling is used for the first 'SamplingEpoch' epochs, afterwards, all events are taken for training]
SamplingImportance: "1.000000e+00" [ The sampling weights of events in epochs which successful (worse estimator than before) are multiplied with SamplingImportance, else they are divided.]
SamplingTraining: "True" [The training sample is sampled]
SamplingTesting: "False" [The testing sample is sampled]
ResetStep: "50" [How often BFGS should reset history]
Tau: "3.000000e+00" [LineSearch "size step"]
BPMode: "sequential" [Back-propagation learning mode: sequential or batch]
BatchSize: "-1" [Batch size: number of events/batch, only set if in Batch Mode, -1 for BatchSize=number_of_events]
ConvergenceImprove: "1.000000e-30" [Minimum improvement which counts as improvement (<0 means automatic convergence check is turned off)]
ConvergenceTests: "-1" [Number of steps (without improvement) required for convergence (<0 means automatic convergence check is turned off)]
UseRegulator: "False" [Use regulator to avoid over-training]
UpdateLimit: "10000" [Maximum times of regulator update]
CalculateErrors: "False" [Calculates inverse Hessian matrix at the end of the training to be able to calculate the uncertainties of an MVA value]
WeightRange: "1.000000e+00" [Take the events for the estimator calculations from small deviations from the desired value to large deviations only over the weight range]
##


#VAR -*-*-*-*-*-*-*-*-*-*-*-* variables *-*-*-*-*-*-*-*-*-*-*-*-

NVar 4
var1                          var1                          var1                          var1                                                            'F'    [-4.05916023254,3.58076572418]
var2                          var2                          var2                          Variable 2                                                      'F'    [-3.68905711174,3.78774046898]
var3                          var3                          var3                          Variable 3                    units                             'F'    [-3.61478614807,4.56402540207]
var4                          var4                          var4                          Variable 4                    units                             'F'    [-4.84856987,5.04116535187]
NSpec 0


============================================================================ */

#include <array>
#include <vector>
#include <cmath>
#include <string>
#include <iostream>

#ifndef IClassifierReader__def
#define IClassifierReader__def

class IClassifierReader {

 public:

   // constructor
   IClassifierReader() : fStatusIsClean( true ) {}
   virtual ~IClassifierReader() {}

   // return classifier response
   virtual double GetMvaValue( const std::vector<double>& inputValues ) const = 0;

   // returns classifier status
   bool IsStatusClean() const { return fStatusIsClean; }

 protected:

   bool fStatusIsClean;
};

#endif

class ReadMLP : public IClassifierReader {

 public:

   // constructor
   ReadMLP( std::vector<std::string>& theInputVars ) 
      : IClassifierReader(),
        fClassName( "ReadMLP" ),
        fNvars( 4 ),
        fIsNormalised( false )
   {      
      // the training input variables
      const char* inputVars[] = { "var1", "var2", "var3", "var4" };

      // sanity checks
      if (theInputVars.size() <= 0) {
         std::cout << "Problem in class \"" << fClassName << "\": empty input vector" << std::endl;
         fStatusIsClean = false;
      }

      if (theInputVars.size() != fNvars) {
         std::cout << "Problem in class \"" << fClassName << "\": mismatch in number of input values: "
                   << theInputVars.size() << " != " << fNvars << std::endl;
         fStatusIsClean = false;
      }

      // validate input variables
      for (size_t ivar = 0; ivar < theInputVars.size(); ivar++) {
         if (theInputVars[ivar] != inputVars[ivar]) {
            std::cout << "Problem in class \"" << fClassName << "\": mismatch in input variable names" << std::endl
                      << " for variable [" << ivar << "]: " << theInputVars[ivar].c_str() << " != " << inputVars[ivar] << std::endl;
            fStatusIsClean = false;
         }
      }

      // initialize min and max vectors (for normalisation)
      fVmin[0] = 0;
      fVmax[0] = 0;
      fVmin[1] = 0;
      fVmax[1] = 0;
      fVmin[2] = 0;
      fVmax[2] = 0;
      fVmin[3] = 0;
      fVmax[3] = 0;

      // initialize input variable types
      fType[0] = 'F';
      fType[1] = 'F';
      fType[2] = 'F';
      fType[3] = 'F';

      // initialize constants
      Initialize();

   }

   // destructor
   virtual ~ReadMLP() {
      Clear(); // method-specific
   }

   // the classifier response
   // "inputValues" is a vector of input values in the same order as the 
   // variables given to the constructor
   double GetMvaValue( const std::vector<double>& inputValues ) const;

 private:

   // method-specific destructor
   void Clear();

   // common member variables
   const char* fClassName;

   const size_t fNvars;
   size_t GetNvar()           const { return fNvars; }
   char   GetType( int ivar ) const { return fType[ivar]; }

   // normalisation of input variables
   const bool fIsNormalised;
   bool IsNormalised() const { return fIsNormalised; }
   double fVmin[4];
   double fVmax[4];
   double NormVariable( double x, double xmin, double xmax ) const {
      // normalise to output range: [-1, 1]
      return 2*(x - xmin)/(xmax - xmin) - 1.0;
   }

   // type of input variable: 'F' or 'I'
   char   fType[4];

   // initialize internal variables
   void Initialize();
   double GetMvaValue__( const std::vector<double>& inputValues ) const;

   // private members (method specific)

   double ActivationFnc(double x) const;
   double OutputActivationFnc(double x) const;

   int fLayers;
   int fLayerSize[4];
   double fWeightMatrix0to1[10][5];   // weight matrix from layer 0 to 1
   double fWeightMatrix1to2[6][10];   // weight matrix from layer 1 to 2
   double fWeightMatrix2to3[4][6];   // weight matrix from layer 2 to 3

};

inline void ReadMLP::Initialize()
{
   // build network structure
   fLayers = 4;
   fLayerSize[0] = 5;
   fLayerSize[1] = 10;
   fLayerSize[2] = 6;
   fLayerSize[3] = 4;
   // weight matrix from layer 0 to 1
   fWeightMatrix0to1[0][0] = -2.56827151871145;
   fWeightMatrix0to1[1][0] = 1.9675350707796;
   fWeightMatrix0to1[2][0] = 0.0945211430434372;
   fWeightMatrix0to1[3][0] = -0.534652842263711;
   fWeightMatrix0to1[4][0] = -1.20582392866552;
   fWeightMatrix0to1[5][0] = -1.15295712224926;
   fWeightMatrix0to1[6][0] = 0.062068122332932;
   fWeightMatrix0to1[7][0] = 1.77909967887342;
   fWeightMatrix0to1[8][0] = -0.465417129592263;
   fWeightMatrix0to1[0][1] = -1.14980631440935;
   fWeightMatrix0to1[1][1] = 0.997243902140403;
   fWeightMatrix0to1[2][1] = 0.23295461257945;
   fWeightMatrix0to1[3][1] = 0.597331762575396;
   fWeightMatrix0to1[4][1] = -1.07812121147322;
   fWeightMatrix0to1[5][1] = -0.904010711693679;
   fWeightMatrix0to1[6][1] = 0.763987509344495;
   fWeightMatrix0to1[7][1] = -0.42730495957135;
   fWeightMatrix0to1[8][1] = 1.98468529543297;
   fWeightMatrix0to1[0][2] = 0.255102649182503;
   fWeightMatrix0to1[1][2] = 1.33042896557683;
   fWeightMatrix0to1[2][2] = -1.01135742504311;
   fWeightMatrix0to1[3][2] = -0.163148696405109;
   fWeightMatrix0to1[4][2] = -0.874674280440262;
   fWeightMatrix0to1[5][2] = -0.605186137715753;
   fWeightMatrix0to1[6][2] = -1.29503636367852;
   fWeightMatrix0to1[7][2] = 0.0166465979160239;
   fWeightMatrix0to1[8][2] = -0.0838533085354363;
   fWeightMatrix0to1[0][3] = -0.298765733434783;
   fWeightMatrix0to1[1][3] = -3.61669821007974;
   fWeightMatrix0to1[2][3] = 0.230635988926395;
   fWeightMatrix0to1[3][3] = 1.41124122416393;
   fWeightMatrix0to1[4][3] = 2.50578787969626;
   fWeightMatrix0to1[5][3] = 1.24044263856194;
   fWeightMatrix0to1[6][3] = 0.444402811598472;
   fWeightMatrix0to1[7][3] = -0.0135466357087942;
   fWeightMatrix0to1[8][3] = 0.0114644267404459;
   fWeightMatrix0to1[0][4] = -2.80469640642522;
   fWeightMatrix0to1[1][4] = 2.03677268622447;
   fWeightMatrix0to1[2][4] = -1.2252550607933;
   fWeightMatrix0to1[3][4] = -0.400363817658847;
   fWeightMatrix0to1[4][4] = 1.74778853696658;
   fWeightMatrix0to1[5][4] = 3.02549270492086;
   fWeightMatrix0to1[6][4] = 2.62632004739209;
   fWeightMatrix0to1[7][4] = -1.05661183010273;
   fWeightMatrix0to1[8][4] = -2.27738442555204;
   // weight matrix from layer 1 to 2
   fWeightMatrix1to2[0][0] = -1.22570337882136;
   fWeightMatrix1to2[1][0] = -0.0931186540811387;
   fWeightMatrix1to2[2][0] = 0.0272542341322508;
   fWeightMatrix1to2[3][0] = 2.20635629364256;
   fWeightMatrix1to2[4][0] = 0.0433728010577296;
   fWeightMatrix1to2[0][1] = 0.273962369245395;
   fWeightMatrix1to2[1][1] = 0.0930582916641537;
   fWeightMatrix1to2[2][1] = -1.19664292914285;
   fWeightMatrix1to2[3][1] = 0.157749723688285;
   fWeightMatrix1to2[4][1] = -4.93727193311647;
   fWeightMatrix1to2[0][2] = -0.715714091172125;
   fWeightMatrix1to2[1][2] = 0.867788095978986;
   fWeightMatrix1to2[2][2] = -1.96198302938044;
   fWeightMatrix1to2[3][2] = -0.464404359512948;
   fWeightMatrix1to2[4][2] = 0.0595675643343206;
   fWeightMatrix1to2[0][3] = 0.758204927770786;
   fWeightMatrix1to2[1][3] = -2.33053777721601;
   fWeightMatrix1to2[2][3] = -0.0596569719756188;
   fWeightMatrix1to2[3][3] = -0.437338637677644;
   fWeightMatrix1to2[4][3] = 0.166445159553785;
   fWeightMatrix1to2[0][4] = 0.254272880836908;
   fWeightMatrix1to2[1][4] = -1.35815091193781;
   fWeightMatrix1to2[2][4] = -0.364973793826042;
   fWeightMatrix1to2[3][4] = 0.461639459085397;
   fWeightMatrix1to2[4][4] = 2.61649708222458;
   fWeightMatrix1to2[0][5] = -0.607652087990045;
   fWeightMatrix1to2[1][5] = -1.59745372261442;
   fWeightMatrix1to2[2][5] = -0.497497443898413;
   fWeightMatrix1to2[3][5] = 1.55870109321722;
   fWeightMatrix1to2[4][5] = 1.76169097043581;
   fWeightMatrix1to2[0][6] = 0.636994330715913;
   fWeightMatrix1to2[1][6] = -0.437626501220008;
   fWeightMatrix1to2[2][6] = -1.64640596581009;
   fWeightMatrix1to2[3][6] = 1.81557603834963;
   fWeightMatrix1to2[4][6] = 0.213460988352797;
   fWeightMatrix1to2[0][7] = 2.87318117712971;
   fWeightMatrix1to2[1][7] = -0.3992286418755;
   fWeightMatrix1to2[2][7] = -0.0828774633250926;
   fWeightMatrix1to2[3][7] = -2.32908344616873;
   fWeightMatrix1to2[4][7] = -0.00287196406204667;
   fWeightMatrix1to2[0][8] = 3.18538225792951;
   fWeightMatrix1to2[1][8] = 0.00204371301044828;
   fWeightMatrix1to2[2][8] = 0.00883819846358683;
   fWeightMatrix1to2[3][8] = -3.25311571046186;
   fWeightMatrix1to2[4][8] = 0.0863656137054146;
   fWeightMatrix1to2[0][9] = -2.58613271127948;
   fWeightMatrix1to2[1][9] = -1.61491969085433;
   fWeightMatrix1to2[2][9] = 1.46349048563992;
   fWeightMatrix1to2[3][9] = -3.52557440843157;
   fWeightMatrix1to2[4][9] = 0.774477596633972;
   // weight matrix from layer 2 to 3
   fWeightMatrix2to3[0][0] = 1.00887795674738;
   fWeightMatrix2to3[1][0] = 0.901934446106593;
   fWeightMatrix2to3[2][0] = 0.0103134082552144;
   fWeightMatrix2to3[3][0] = -4.15846758201171;
   fWeightMatrix2to3[0][1] = -1.76663219370744;
   fWeightMatrix2to3[1][1] = -3.16703807597721;
   fWeightMatrix2to3[2][1] = -0.721676163411028;
   fWeightMatrix2to3[3][1] = 2.88184594216196;
   fWeightMatrix2to3[0][2] = 1.05464986697278;
   fWeightMatrix2to3[1][2] = -1.33953128311584;
   fWeightMatrix2to3[2][2] = 2.06932178843474;
   fWeightMatrix2to3[3][2] = -0.566502440872972;
   fWeightMatrix2to3[0][3] = 1.1754576459947;
   fWeightMatrix2to3[1][3] = 1.04229738415628;
   fWeightMatrix2to3[2][3] = 0.409096508127509;
   fWeightMatrix2to3[3][3] = -4.91383178074721;
   fWeightMatrix2to3[0][4] = 2.89032187783364;
   fWeightMatrix2to3[1][4] = -0.640305503795561;
   fWeightMatrix2to3[2][4] = -3.0600188888794;
   fWeightMatrix2to3[3][4] = 0.412788618104668;
   fWeightMatrix2to3[0][5] = 0.441316337386854;
   fWeightMatrix2to3[1][5] = -0.0238371766698862;
   fWeightMatrix2to3[2][5] = 0.903246146264539;
   fWeightMatrix2to3[3][5] = 0.705588849148349;
}

inline double ReadMLP::GetMvaValue__( const std::vector<double>& inputValues ) const
{
   if (inputValues.size() != (unsigned int)fLayerSize[0]-1) {
      std::cout << "Input vector needs to be of size " << fLayerSize[0]-1 << std::endl;
      return 0;
   }

   std::array<double, 5> fWeights0 {{}};
   std::array<double, 10> fWeights1 {{}};
   std::array<double, 6> fWeights2 {{}};
   std::array<double, 4> fWeights3 {{}};
   fWeights0.back() = 1.;
   fWeights1.back() = 1.;
   fWeights2.back() = 1.;

   for (int i=0; i<fLayerSize[0]-1; i++)
      fWeights0[i]=inputValues[i];

   // layer 0 to 1
   for (int o=0; o<fLayerSize[1]-1; o++) {
      for (int i=0; i<fLayerSize[0]; i++) {
         double inputVal = fWeightMatrix0to1[o][i] * fWeights0[i];
         fWeights1[o] += inputVal;
      } // loop over i
      fWeights1[o] = ActivationFnc(fWeights1[o]);
   } // loop over o
   // layer 1 to 2
   for (int o=0; o<fLayerSize[2]-1; o++) {
      for (int i=0; i<fLayerSize[1]; i++) {
         double inputVal = fWeightMatrix1to2[o][i] * fWeights1[i];
         fWeights2[o] += inputVal;
      } // loop over i
      fWeights2[o] = ActivationFnc(fWeights2[o]);
   } // loop over o
   // layer 2 to 3
   for (int o=0; o<fLayerSize[3]; o++) {
      for (int i=0; i<fLayerSize[2]; i++) {
         double inputVal = fWeightMatrix2to3[o][i] * fWeights2[i];
         fWeights3[o] += inputVal;
      } // loop over i
      fWeights3[o] = OutputActivationFnc(fWeights3[o]);
   } // loop over o

   return fWeights3[0];
}

double ReadMLP::ActivationFnc(double x) const {
   // hyperbolic tan
   return tanh(x);
}
double ReadMLP::OutputActivationFnc(double x) const {
   // identity
   return x;
}
   
// Clean up
inline void ReadMLP::Clear() 
{
}
   inline double ReadMLP::GetMvaValue( const std::vector<double>& inputValues ) const
   {
      // classifier response value
      double retval = 0;

      // classifier response, sanity check first
      if (!IsStatusClean()) {
         std::cout << "Problem in class \"" << fClassName << "\": cannot return classifier response"
                   << " because status is dirty" << std::endl;
         retval = 0;
      }
      else {
         if (IsNormalised()) {
            // normalise variables
            std::vector<double> iV;
            iV.reserve(inputValues.size());
            int ivar = 0;
            for (std::vector<double>::const_iterator varIt = inputValues.begin();
                 varIt != inputValues.end(); varIt++, ivar++) {
               iV.push_back(NormVariable( *varIt, fVmin[ivar], fVmax[ivar] ));
            }
            retval = GetMvaValue__( iV );
         }
         else {
            retval = GetMvaValue__( inputValues );
         }
      }

      return retval;
   }