Dataset for Sideband Subtraction Wraps a dataset and provides each data-point with a new weight. More...

#include <DataDriven.h>

Inheritance diagram for SidebandDataset:

Collaboration diagram for SidebandDataset:

[legend]

Public Member Functions
	SidebandDataset (const GeneralOptions &general_options, Dataset &dataset, Dataset &mc_dataset, const std::string &sideband_variable)
	Constructs a new SidebandDataset.

virtual unsigned int	getNumberOfFeatures () const override
	Returns the number of features in this dataset.

virtual unsigned int	getNumberOfSpectators () const override
	Returns the number of features in this dataset.

virtual unsigned int	getNumberOfEvents () const override
	Returns the number of events in this dataset.

virtual std::vector< float >	getFeature (unsigned int iFeature) override
	Returns all values of one feature in a std::vector<float>

virtual std::vector< float >	getSpectator (unsigned int iSpectator) override
	Returns all values of one spectator in a std::vector<float>

virtual void	loadEvent (unsigned int event) override
	Load the event number iEvent.

virtual float	getSignalFraction ()
	Returns the signal fraction of the whole sample.

virtual unsigned int	getFeatureIndex (const std::string &feature)
	Return index of feature with the given name.

virtual unsigned int	getSpectatorIndex (const std::string &spectator)
	Return index of spectator with the given name.

virtual std::vector< float >	getWeights ()
	Returns all weights.

virtual std::vector< float >	getTargets ()
	Returns all targets.

virtual std::vector< bool >	getSignals ()
	Returns all is Signals.

Public Attributes
GeneralOptions	m_general_options
	GeneralOptions passed to this dataset.

std::vector< float >	m_input
	Contains all feature values of the currently loaded event.

std::vector< float >	m_spectators
	Contains all spectators values of the currently loaded event.

float	m_weight
	Contains the weight of the currently loaded event.

float	m_target
	Contains the target value of the currently loaded event.

bool	m_isSignal
	Defines if the currently loaded event is signal or background.

Private Attributes
Dataset &	m_dataset
	Wrapped dataset.

int	m_spectator_index
	spectator containing the sideband variable

double	m_signal_weight
	the weight for signal events

double	m_background_weight
	the weight for background events

double	m_negative_signal_weight
	the weight for negative signal events

Detailed Description

Dataset for Sideband Subtraction Wraps a dataset and provides each data-point with a new weight.

Definition at line 104 of file DataDriven.h.

Constructor & Destructor Documentation

◆ SidebandDataset()

SidebandDataset	(	const GeneralOptions &	general_options,
		Dataset &	dataset,
		Dataset &	mc_dataset,
		const std::string &	sideband_variable
	)

Constructs a new SidebandDataset.

Parameters

general_options	shared options defining the dataset (variables, ...)
dataset	containing the data-points
mc_dataset	containing mc
sideband_variable	defining the sideband regions

Definition at line 38 of file DataDriven.cc.

                                                                         : Dataset(general_options), m_dataset(dataset)
    {
 
      m_spectator_index = dataset.getSpectatorIndex(sideband_variable);
      int mc_spectator_index = mc_dataset.getSpectatorIndex(sideband_variable);
 
      double total_signal_mc = 0.0;
      double total_mc = 0.0;
      double sum_signal_sr = 0.0;
      double sum_sr = 0.0;
      double sum_signal_br = 0.0;
      double sum_br = 0.0;
      double sum_signal_nr = 0.0;
      double sum_nr = 0.0;
 
      for (unsigned int iEvent = 0; iEvent < mc_dataset.getNumberOfEvents(); ++iEvent) {
        mc_dataset.loadEvent(iEvent);
        if (mc_dataset.m_isSignal)
          total_signal_mc += mc_dataset.m_weight;
        total_mc += mc_dataset.m_weight;
        if (mc_dataset.m_spectators[mc_spectator_index] == 1.0) {
          if (mc_dataset.m_isSignal)
            sum_signal_sr += mc_dataset.m_weight;
          sum_sr += mc_dataset.m_weight;
        } else if (mc_dataset.m_spectators[mc_spectator_index] == 2.0) {
          if (mc_dataset.m_isSignal)
            sum_signal_br += mc_dataset.m_weight;
          sum_br += mc_dataset.m_weight;
        } else if (mc_dataset.m_spectators[mc_spectator_index] == 3.0) {
          if (mc_dataset.m_isSignal)
            sum_signal_nr += mc_dataset.m_weight;
          sum_nr += mc_dataset.m_weight;
        }
      }
 
      double total_data = 0.0;
      double sum_data_sr = 0.0;
      double sum_data_br = 0.0;
      double sum_data_nr = 0.0;
 
      for (unsigned int iEvent = 0; iEvent < dataset.getNumberOfEvents(); ++iEvent) {
        dataset.loadEvent(iEvent);
        total_data += dataset.m_weight;
        if (dataset.m_spectators[m_spectator_index] == 1.0) {
          sum_data_sr += dataset.m_weight;
        } else if (dataset.m_spectators[m_spectator_index] == 2.0) {
          sum_data_br += dataset.m_weight;
        } else if (dataset.m_spectators[m_spectator_index] == 3.0) {
          sum_data_nr += dataset.m_weight;
        }
      }
 
      if (sum_signal_br / sum_br > 0.01) {
        B2WARNING("The background region you defined in the sideband subtraction contains more than 1% signal");
      }
      if (sum_signal_nr / sum_nr > 0.01) {
        B2WARNING("The negative signal region you defined in the sideband subtraction contains more than 1% signal");
      }
 
      if (sum_data_sr - sum_signal_sr < 0) {
        B2ERROR("There is less data in the signal region than the expected amount of signal events in the signal region estimated from MC.");
      }
 
      if (total_data - total_signal_mc < 0) {
        B2ERROR("There is less data than the expected amount of signal events estimated from MC.");
      }
 
      // We assume the number of signal events is correctly described in mc
      // Everything else (like the background) we take from the data sample
 
      // So Signal events in the signal region receive weight 1
      m_signal_weight = 1.0;
 
      // The background is scaled so that it corresponds to the total background in the whole sample
      m_background_weight = (total_data - total_signal_mc) / sum_data_br;
 
      // The negative signal is scaled so it corresponds to the expected background in the signal region:
      // Background Events in Signal Region in Data = Total Events in Signal Region In Data - Signal Events in Signal Region in MC
      m_negative_signal_weight = - (sum_data_sr - sum_signal_sr) / sum_data_nr;
 
      B2INFO("Data " << total_data << " " << sum_data_sr << " " << sum_data_br << " " << sum_data_nr);
      B2INFO("MC " << total_mc << " " << sum_sr << " " << sum_br << " " << sum_nr);
      B2INFO("MC (signal)" << total_signal_mc << " " << sum_signal_sr << " " << sum_signal_br << " " << sum_signal_nr);
      B2INFO("Sideband Subtraction: Signal Weight " << m_signal_weight << " Background Weight " << m_background_weight <<
             " Negative Signal Weight " << m_negative_signal_weight);
 
    }

Member Function Documentation

◆ getFeature()

virtual std::vector< float > getFeature ( unsigned int iFeature )

inlineoverridevirtual

Returns all values of one feature in a std::vector<float>

Parameters

iFeature the position of the feature to return

Reimplemented from Dataset.

Definition at line 135 of file DataDriven.h.

135{ return m_dataset.getFeature(iFeature); }

Belle2::MVA::Dataset::getFeature

virtual std::vector< float > getFeature(unsigned int iFeature)

Returns all values of one feature in a std::vector<float>

Definition: Dataset.cc:74

◆ getFeatureIndex()

unsigned int getFeatureIndex ( const std::string & feature )

virtualinherited

Return index of feature with the given name.

Parameters

feature name of the feature

Definition at line 50 of file Dataset.cc.

    {
 
      auto it = std::find(m_general_options.m_variables.begin(), m_general_options.m_variables.end(), feature);
      if (it == m_general_options.m_variables.end()) {
        B2ERROR("Unknown feature named " << feature);
        return 0;
      }
      return std::distance(m_general_options.m_variables.begin(), it);
 
    }

◆ getNumberOfEvents()

virtual unsigned int getNumberOfEvents ( ) const

inlineoverridevirtual

Returns the number of events in this dataset.

Implements Dataset.

Definition at line 129 of file DataDriven.h.

129{ return m_dataset.getNumberOfEvents(); };

Belle2::MVA::Dataset::getNumberOfEvents

virtual unsigned int getNumberOfEvents() const =0

Returns the number of events in this dataset.

◆ getNumberOfFeatures()

virtual unsigned int getNumberOfFeatures ( ) const

inlineoverridevirtual

Returns the number of features in this dataset.

Implements Dataset.

Definition at line 119 of file DataDriven.h.

119{ return m_dataset.getNumberOfFeatures(); }

Belle2::MVA::Dataset::getNumberOfFeatures

virtual unsigned int getNumberOfFeatures() const =0

Returns the number of features in this dataset.

◆ getNumberOfSpectators()

virtual unsigned int getNumberOfSpectators ( ) const

inlineoverridevirtual

Returns the number of features in this dataset.

Implements Dataset.

Definition at line 124 of file DataDriven.h.

124{ return m_dataset.getNumberOfSpectators(); }

Belle2::MVA::Dataset::getNumberOfSpectators

virtual unsigned int getNumberOfSpectators() const =0

Returns the number of spectators in this dataset.

◆ getSignalFraction()

float getSignalFraction ( )

virtualinherited

Returns the signal fraction of the whole sample.

Reimplemented in SPlotDataset.

Definition at line 35 of file Dataset.cc.

    {
 
      double signal_weight_sum = 0;
      double weight_sum = 0;
      for (unsigned int i = 0; i < getNumberOfEvents(); ++i) {
        loadEvent(i);
        weight_sum += m_weight;
        if (m_isSignal)
          signal_weight_sum += m_weight;
      }
      return signal_weight_sum / weight_sum;
 
    }

◆ getSignals()

std::vector< bool > getSignals ( )

virtualinherited

Returns all is Signals.

Reimplemented in ReweightingDataset.

Definition at line 122 of file Dataset.cc.

    {
 
      std::vector<bool> result(getNumberOfEvents());
      for (unsigned int iEvent = 0; iEvent < getNumberOfEvents(); ++iEvent) {
        loadEvent(iEvent);
        result[iEvent] = m_isSignal;
      }
      return result;
 
    }

◆ getSpectator()

virtual std::vector< float > getSpectator ( unsigned int iSpectator )

inlineoverridevirtual

Returns all values of one spectator in a std::vector<float>

Parameters

iSpectator the position of the feature to return

Reimplemented from Dataset.

Definition at line 141 of file DataDriven.h.

141{ return m_dataset.getSpectator(iSpectator); }

Belle2::MVA::Dataset::getSpectator

virtual std::vector< float > getSpectator(unsigned int iSpectator)

Returns all values of one spectator in a std::vector<float>

Definition: Dataset.cc:86

◆ getSpectatorIndex()

unsigned int getSpectatorIndex ( const std::string & spectator )

virtualinherited

Return index of spectator with the given name.

Parameters

spectator name of the spectator

Definition at line 62 of file Dataset.cc.

    {
 
      auto it = std::find(m_general_options.m_spectators.begin(), m_general_options.m_spectators.end(), spectator);
      if (it == m_general_options.m_spectators.end()) {
        B2ERROR("Unknown spectator named " << spectator);
        return 0;
      }
      return std::distance(m_general_options.m_spectators.begin(), it);
 
    }

◆ getTargets()

std::vector< float > getTargets ( )

virtualinherited

Returns all targets.

Reimplemented in RegressionDataSet, and ReweightingDataset.

Definition at line 110 of file Dataset.cc.

    {
 
      std::vector<float> result(getNumberOfEvents());
      for (unsigned int iEvent = 0; iEvent < getNumberOfEvents(); ++iEvent) {
        loadEvent(iEvent);
        result[iEvent] = m_target;
      }
      return result;
 
    }

◆ getWeights()

std::vector< float > getWeights ( )

virtualinherited

Returns all weights.

Reimplemented in ROOTDataset, RegressionDataSet, and ReweightingDataset.

Definition at line 98 of file Dataset.cc.

    {
 
      std::vector<float> result(getNumberOfEvents());
      for (unsigned int iEvent = 0; iEvent < getNumberOfEvents(); ++iEvent) {
        loadEvent(iEvent);
        result[iEvent] = m_weight;
      }
      return result;
 
    }

◆ loadEvent()

void loadEvent ( unsigned int event )

overridevirtual

Load the event number iEvent.

Parameters

event event number to load

Implements Dataset.

Definition at line 127 of file DataDriven.cc.

    {
      m_dataset.loadEvent(event);
      m_input = m_dataset.m_input;
      m_spectators = m_dataset.m_spectators;
      m_weight = m_dataset.m_weight;
      if (m_spectators[m_spectator_index] == 1.0) {
        m_isSignal = true;
        m_target = 1.0;
        m_weight *= m_signal_weight;
      } else if (m_spectators[m_spectator_index] == 2.0) {
        m_isSignal = false;
        m_target = 0.0;
        m_weight *= m_background_weight;
      } else if (m_spectators[m_spectator_index] == 3.0) {
        m_isSignal = true;
        m_target = 1.0;
        m_weight *= m_negative_signal_weight;
      } else {
        m_isSignal = false;
        m_target = 0.0;
        m_weight = 0.0;
      }
    }

Definition at line 126 of file Dataset.h.

◆ m_weight

float m_weight

inherited

Contains the weight of the currently loaded event.

Definition at line 125 of file Dataset.h.

The documentation for this class was generated from the following files:

mva/utility/include/DataDriven.h
mva/utility/src/DataDriven.cc

Public Member Functions

Public Attributes

Private Attributes

Detailed Description

Constructor & Destructor Documentation

◆ SidebandDataset()

Member Function Documentation

◆ getFeature()

◆ getFeatureIndex()

◆ getNumberOfEvents()

◆ getNumberOfFeatures()

◆ getNumberOfSpectators()

◆ getSignalFraction()

◆ getSignals()

◆ getSpectator()

◆ getSpectatorIndex()

◆ getTargets()

◆ getWeights()

◆ loadEvent()

Member Data Documentation

◆ m_background_weight

◆ m_dataset

◆ m_general_options

◆ m_input

◆ m_isSignal

◆ m_negative_signal_weight

◆ m_signal_weight

◆ m_spectator_index

◆ m_spectators

◆ m_target

◆ m_weight