emtf::Node Class Reference

#include <Node.h>

Public Member Functions
void	calcOptimumSplit ()
void	filterEventsToDaughters ()
Node *	filterEventToDaughter (Event *e)
double	getAvgError ()
double	getErrorReduction ()
std::vector< std::vector< Event * > > &	getEvents ()
double	getFitValue ()
Node *	getLeftDaughter ()
std::string	getName ()
int	getNumEvents ()
Node *	getParent ()
Node *	getRightDaughter ()
double	getSplitValue ()
int	getSplitVariable ()
double	getTotalError ()
void	listEvents ()
	Node ()
	Node (std::string cName)
	Node (Node &&)=default
	Node (const Node &)=delete
Node &	operator= (const Node &)=delete
void	setAvgError (double sAvgError)
void	setErrorReduction (double sErrorReduction)
void	setEvents (std::vector< std::vector< Event *> > &sEvents)
void	setFitValue (double sFitValue)
void	setLeftDaughter (Node *sLeftDaughter)
void	setName (std::string sName)
void	setNumEvents (int sNumEvents)
void	setParent (Node *sParent)
void	setRightDaughter (Node *sLeftDaughter)
void	setSplitValue (double sSplitValue)
void	setSplitVariable (int sSplitVar)
void	setTotalError (double sTotalError)
void	theMiracleOfChildBirth ()
	~Node ()

Private Attributes
double	avgError
double	errorReduction
std::vector< std::vector< Event * > >	events
double	fitValue
Node *	leftDaughter
std::string	name
int	numEvents
Node *	parent
Node *	rightDaughter
double	splitValue
int	splitVariable
double	totalError

Detailed Description

Definition at line 12 of file Node.h.

Constructor & Destructor Documentation

Node() [1/4]

Node::Node

(

)

Definition at line 32 of file Node.cc.

References avgError, errorReduction, leftDaughter, name, parent, rightDaughter, splitValue, splitVariable, and totalError.

Referenced by theMiracleOfChildBirth().

           {
  name = "";
  leftDaughter = nullptr;
  rightDaughter = nullptr;
  parent = nullptr;
  splitValue = -99;
  splitVariable = -1;
  avgError = -1;
  totalError = -1;
  errorReduction = -1;
}

Node() [2/4]

Node::Node

(

std::string

cName

)

Definition at line 44 of file Node.cc.

References avgError, errorReduction, leftDaughter, name, parent, rightDaughter, splitValue, splitVariable, and totalError.

                          {
  name = cName;
  leftDaughter = nullptr;
  rightDaughter = nullptr;
  parent = nullptr;
  splitValue = -99;
  splitVariable = -1;
  avgError = -1;
  totalError = -1;
  errorReduction = -1;
}

~Node()

Node::~Node

(

)

Definition at line 60 of file Node.cc.

References leftDaughter, and rightDaughter.

            {
  // Recursively delete all nodes in the tree.
  if (leftDaughter)
    delete leftDaughter;
  if (rightDaughter)
    delete rightDaughter;
}

Node() [3/4]

emtf::Node::Node ( Node &&  )

default

Node() [4/4]

emtf::Node::Node ( const Node &  )

delete

Member Function Documentation

calcOptimumSplit()

void Node::calcOptimumSplit

(

)

Definition at line 143 of file Node.cc.

References avgError, data, errorReduction, events, fitValue, mps_fire::i, numEvents, splitValue, splitVariable, SUM, filterCSVwithJSON::target, totalError, and findQualityFiles::v.

Referenced by emtf::Tree::buildTree().

                            {
  // Determines the split variable and split point which would most reduce the error for the given node (region).
  // In the process we calculate the fitValue and Error. The general aglorithm is based upon  Luis Torgo's thesis.
  // Check out the reference for a more in depth outline. This part is chapter 3.

  // Intialize some variables.
  double bestSplitValue = 0;
  int bestSplitVariable = -1;
  double bestErrorReduction = -1;

  double SUM = 0;
  double SSUM = 0;
  numEvents = events[0].size();

  double candidateErrorReduction = -1;

  // Calculate the sum of the target variables and the sum of
  // the target variables squared. We use these later.
  for (unsigned int i = 0; i < events[0].size(); i++) {
    double target = events[0][i]->data[0];
    SUM += target;
    SSUM += target * target;
  }

  unsigned int numVars = events.size();

  // Calculate the best split point for each variable
  for (unsigned int variableToCheck = 1; variableToCheck < numVars; variableToCheck++) {
    // The sum of the target variables in the left, right nodes
    double SUMleft = 0;
    double SUMright = SUM;

    // The number of events in the left, right nodes
    int nleft = 1;
    int nright = events[variableToCheck].size() - 1;

    int candidateSplitVariable = variableToCheck;

    std::vector<Event*>& v = events[variableToCheck];

    // Find the best split point for this variable
    for (unsigned int i = 1; i < v.size(); i++) {
      // As the candidate split point interates, the number of events in the
      // left/right node increases/decreases and SUMleft/right increases/decreases.

      SUMleft = SUMleft + v[i - 1]->data[0];
      SUMright = SUMright - v[i - 1]->data[0];

      // No need to check the split point if x on both sides is equal
      if (v[i - 1]->data[candidateSplitVariable] < v[i]->data[candidateSplitVariable]) {
        // Finding the maximum error reduction for Least Squares boils down to maximizing
        // the following statement.
        candidateErrorReduction = SUMleft * SUMleft / nleft + SUMright * SUMright / nright - SUM * SUM / numEvents;
        //                std::cout << "candidateErrorReduction= " << candidateErrorReduction << std::endl << std::endl;

        // if the new candidate is better than the current best, then we have a new overall best.
        if (candidateErrorReduction > bestErrorReduction) {
          bestErrorReduction = candidateErrorReduction;
          bestSplitValue = (v[i - 1]->data[candidateSplitVariable] + v[i]->data[candidateSplitVariable]) / 2;
          bestSplitVariable = candidateSplitVariable;
        }
      }

      nright = nright - 1;
      nleft = nleft + 1;
    }
  }

  // Store the information gained from our computations.

  // The fit value is the average for least squares.
  fitValue = SUM / numEvents;
  //    std::cout << "fitValue= " << fitValue << std::endl;

  // n*[ <y^2>-k^2 ]
  totalError = SSUM - SUM * SUM / numEvents;
  //    std::cout << "totalError= " << totalError << std::endl;

  // [ <y^2>-k^2 ]
  avgError = totalError / numEvents;
  //    std::cout << "avgError= " << avgError << std::endl;

  errorReduction = bestErrorReduction;
  //    std::cout << "errorReduction= " << errorReduction << std::endl;

  splitVariable = bestSplitVariable;
  //    std::cout << "splitVariable= " << splitVariable << std::endl;

  splitValue = bestSplitValue;
  //    std::cout << "splitValue= " << splitValue << std::endl;

  //if(bestSplitVariable == -1) std::cout << "splitVar = -1. numEvents = " << numEvents << ". errRed = " << errorReduction << std::endl;
}

filterEventsToDaughters()

void Node::filterEventsToDaughters

(

)

Definition at line 267 of file Node.cc.

References MillePedeFileConverter_cfg::e, events, getEvents(), mps_fire::i, dqmiolumiharvest::j, cmsLHEtoEOSManager::l, leftDaughter, rightDaughter, setNumEvents(), splitValue, splitVariable, and pfDeepBoostedJetPreprocessParams_cfi::sv.

Referenced by emtf::Tree::buildTree(), and emtf::Tree::filterEventsRecursive().

                                   {
  // Keeping sorted copies of the event vectors allows us to save on
  // computation time. That way we don't have to resort the events
  // each time we calculate the splitpoint for a node. We sort them once.
  // Every time we split a node, we simply filter them down correctly
  // preserving the order. This way we have O(n) efficiency instead
  // of O(nlogn) efficiency.

  // Anyways, this function takes events from the parent node
  // and filters an event into the left or right daughter
  // node depending on whether it is < or > the split point
  // for the given split variable.

  unsigned int sv = splitVariable;
  double sp = splitValue;

  Node* left = leftDaughter;
  Node* right = rightDaughter;

  std::vector<std::vector<Event*> > l(events.size());
  std::vector<std::vector<Event*> > r(events.size());

  for (unsigned int i = 0; i < events.size(); i++) {
    for (unsigned int j = 0; j < events[i].size(); j++) {
      Event* e = events[i][j];
      // Prevent out-of-bounds access
      if (sv >= e->data.size())
        continue;
      if (e->data[sv] < sp)
        l[i].push_back(e);
      if (e->data[sv] > sp)
        r[i].push_back(e);
    }
  }

  events = std::vector<std::vector<Event*> >();

  left->getEvents().swap(l);
  right->getEvents().swap(r);

  // Set the number of events in the node.
  left->setNumEvents(left->getEvents()[0].size());
  right->setNumEvents(right->getEvents()[0].size());
}

filterEventToDaughter()

Node * Node::filterEventToDaughter

(

Event *

e

)

Definition at line 314 of file Node.cc.

References MillePedeFileConverter_cfg::e, leftDaughter, rightDaughter, splitValue, splitVariable, and pfDeepBoostedJetPreprocessParams_cfi::sv.

Referenced by emtf::Tree::filterEventRecursive().

                                          {
  // Anyways, this function takes an event from the parent node
  // and filters an event into the left or right daughter
  // node depending on whether it is < or > the split point
  // for the given split variable.

  unsigned int sv = splitVariable;
  double sp = splitValue;

  Node* left = leftDaughter;
  Node* right = rightDaughter;
  Node* nextNode = nullptr;

  // Prevent out-of-bounds access
  if (left == nullptr || right == nullptr || sv >= e->data.size())
    return nullptr;

  if (e->data[sv] < sp)
    nextNode = left;
  if (e->data[sv] >= sp)
    nextNode = right;

  return nextNode;
}

getAvgError()

double Node::getAvgError

(

)

Definition at line 122 of file Node.cc.

References avgError.

Referenced by emtf::Tree::copyFrom().

{ return avgError; }

getErrorReduction()

double Node::getErrorReduction

(

)

Definition at line 80 of file Node.cc.

References errorReduction.

Referenced by emtf::Tree::copyFrom(), and emtf::Tree::rankVariablesRecursive().

{ return errorReduction; }

getEvents()

std::vector< std::vector< Event * > > & Node::getEvents

(

)

Definition at line 132 of file Node.cc.

References events.

Referenced by emtf::Tree::filterEvents(), and filterEventsToDaughters().

{ return events; }

getFitValue()

double Node::getFitValue

(

)

Definition at line 112 of file Node.cc.

References fitValue.

Referenced by emtf::Tree::addXMLAttributes(), emtf::Forest::appendCorrection(), emtf::Tree::copyFrom(), and L1TMuonEndCapForestESProducer::traverse().

{ return fitValue; }

getLeftDaughter()

Node * Node::getLeftDaughter

(

)

Definition at line 86 of file Node.cc.

References leftDaughter.

Referenced by emtf::Tree::buildTree(), emtf::Tree::copyFrom(), emtf::Tree::filterEventsRecursive(), emtf::Tree::findLeafs(), emtf::Tree::getSplitValuesRecursive(), emtf::Tree::loadFromCondPayloadRecursive(), emtf::Tree::loadFromXMLRecursive(), emtf::Tree::rankVariablesRecursive(), emtf::Tree::saveToXMLRecursive(), and L1TMuonEndCapForestESProducer::traverse().

{ return leftDaughter; }

getName()

std::string Node::getName

(

void

)

Definition at line 74 of file Node.cc.

References name.

Referenced by emtf::Tree::copyFrom(), plotting.Plot::draw(), and emtf::Tree::saveToXML().

{ return name; }

getNumEvents()

int Node::getNumEvents

(

)

Definition at line 128 of file Node.cc.

References numEvents.

Referenced by emtf::Tree::calcError(), and emtf::Tree::copyFrom().

{ return numEvents; }

getParent()

Node * Node::getParent

(

)

Definition at line 96 of file Node.cc.

References parent.

{ return parent; }

getRightDaughter()

Node * Node::getRightDaughter

(

)

Definition at line 90 of file Node.cc.

References rightDaughter.

Referenced by emtf::Tree::buildTree(), emtf::Tree::copyFrom(), emtf::Tree::filterEventsRecursive(), emtf::Tree::findLeafs(), emtf::Tree::getSplitValuesRecursive(), emtf::Tree::loadFromCondPayloadRecursive(), emtf::Tree::loadFromXMLRecursive(), emtf::Tree::rankVariablesRecursive(), emtf::Tree::saveToXMLRecursive(), and L1TMuonEndCapForestESProducer::traverse().

{ return rightDaughter; }

getSplitValue()

double Node::getSplitValue

(

)

Definition at line 102 of file Node.cc.

References splitValue.

Referenced by emtf::Tree::addXMLAttributes(), emtf::Tree::copyFrom(), emtf::Tree::getSplitValuesRecursive(), and L1TMuonEndCapForestESProducer::traverse().

{ return splitValue; }

getSplitVariable()

int Node::getSplitVariable

(

)

Definition at line 106 of file Node.cc.

References splitVariable.

Referenced by emtf::Tree::addXMLAttributes(), emtf::Tree::copyFrom(), emtf::Tree::getSplitValuesRecursive(), emtf::Tree::rankVariablesRecursive(), and L1TMuonEndCapForestESProducer::traverse().

{ return splitVariable; }

getTotalError()

double Node::getTotalError

(

)

Definition at line 118 of file Node.cc.

References totalError.

Referenced by emtf::Tree::copyFrom().

{ return totalError; }

listEvents()

void Node::listEvents

(

)

Definition at line 239 of file Node.cc.

References gather_cfg::cout, events, mps_fire::i, and dqmiolumiharvest::j.

                      {
  std::cout << std::endl << "Listing Events... " << std::endl;

  for (unsigned int i = 0; i < events.size(); i++) {
    std::cout << std::endl << "Variable " << i << " vector contents: " << std::endl;
    for (unsigned int j = 0; j < events[i].size(); j++) {
      events[i][j]->outputEvent();
    }
    std::cout << std::endl;
  }
}

operator=()

Node& emtf::Node::operator= ( const Node &  )

delete

setAvgError()

void Node::setAvgError

(

double

sAvgError

)

Definition at line 120 of file Node.cc.

References avgError.

{ avgError = sAvgError; }

setErrorReduction()

void Node::setErrorReduction

(

double

sErrorReduction

)

Definition at line 78 of file Node.cc.

References errorReduction.

{ errorReduction = sErrorReduction; }

setEvents()

void Node::setEvents

(

std::vector< std::vector< Event *> > &

sEvents

)

Definition at line 134 of file Node.cc.

References events, and numEvents.

Referenced by emtf::Tree::Tree().

                                                           {
  events = sEvents;
  numEvents = events[0].size();
}

setFitValue()

void Node::setFitValue

(

double

sFitValue

)

Definition at line 110 of file Node.cc.

References fitValue.

Referenced by emtf::Tree::loadFromCondPayloadRecursive(), and emtf::Tree::loadFromXMLRecursive().

{ fitValue = sFitValue; }

setLeftDaughter()

void Node::setLeftDaughter

(

Node *

sLeftDaughter

)

Definition at line 84 of file Node.cc.

References leftDaughter.

{ leftDaughter = sLeftDaughter; }

setName()

void Node::setName

(

std::string

sName

)

Definition at line 72 of file Node.cc.

References name.

{ name = sName; }

setNumEvents()

void Node::setNumEvents

(

int

sNumEvents

)

Definition at line 126 of file Node.cc.

References numEvents.

Referenced by filterEventsToDaughters().

{ numEvents = sNumEvents; }

setParent()

void Node::setParent

(

Node *

sParent

)

Definition at line 94 of file Node.cc.

References parent.

Referenced by emtf::Tree::copyFrom(), and theMiracleOfChildBirth().

{ parent = sParent; }

setRightDaughter()

void Node::setRightDaughter

(

Node *

sLeftDaughter

)

Definition at line 88 of file Node.cc.

References rightDaughter.

{ rightDaughter = sRightDaughter; }

setSplitValue()

void Node::setSplitValue

(

double

sSplitValue

)

Definition at line 100 of file Node.cc.

References splitValue.

Referenced by emtf::Tree::loadFromCondPayloadRecursive(), and emtf::Tree::loadFromXMLRecursive().

{ splitValue = sSplitValue; }

setSplitVariable()

void Node::setSplitVariable

(

int

sSplitVar

)

Definition at line 104 of file Node.cc.

References splitVariable.

Referenced by emtf::Tree::loadFromCondPayloadRecursive(), and emtf::Tree::loadFromXMLRecursive().

{ splitVariable = sSplitVar; }

setTotalError()

void Node::setTotalError

(

double

sTotalError

)

Definition at line 116 of file Node.cc.

References totalError.

{ totalError = sTotalError; }

theMiracleOfChildBirth()

void Node::theMiracleOfChildBirth

(

)

Definition at line 253 of file Node.cc.

References leftDaughter, name, Node(), rightDaughter, and setParent().

Referenced by emtf::Tree::buildTree(), emtf::Tree::loadFromCondPayloadRecursive(), and emtf::Tree::loadFromXMLRecursive().

                                  {
  // Create Daughter Nodes
  Node* left = new Node(name + " left");
  Node* right = new Node(name + " right");

  // Link the Nodes Appropriately
  leftDaughter = left;
  rightDaughter = right;
  left->setParent(this);
  right->setParent(this);
}

Member Data Documentation

avgError

double emtf::Node::avgError

private

Definition at line 76 of file Node.h.

Referenced by calcOptimumSplit(), getAvgError(), Node(), and setAvgError().

errorReduction

double emtf::Node::errorReduction

private

Definition at line 74 of file Node.h.

Referenced by calcOptimumSplit(), getErrorReduction(), Node(), and setErrorReduction().

events

std::vector<std::vector<Event *> > emtf::Node::events

private

Definition at line 81 of file Node.h.

Referenced by eventsfwlite.Events::__getattr__(), eventsfwlite.Events::__getitem__(), eventsfwlite.Events::__len__(), calcOptimumSplit(), filterEventsToDaughters(), getEvents(), listEvents(), looper.Looper::loop(), looper.Looper::process(), and setEvents().

fitValue

double emtf::Node::fitValue

private

Definition at line 78 of file Node.h.

Referenced by calcOptimumSplit(), getFitValue(), and setFitValue().

leftDaughter

Node* emtf::Node::leftDaughter

private

Definition at line 67 of file Node.h.

Referenced by filterEventsToDaughters(), filterEventToDaughter(), getLeftDaughter(), Node(), setLeftDaughter(), theMiracleOfChildBirth(), and ~Node().

name

std::string emtf::Node::name

private

Definition at line 65 of file Node.h.

Referenced by ElectronMVAID.ElectronMVAID::__call__(), FWLite.ElectronMVAID::__call__(), dirstructure.Directory::__create_pie_image(), DisplayManager.DisplayManager::__del__(), dqm_interfaces.DirID::__eq__(), dirstructure.Directory::__get_full_path(), dirstructure.Comparison::__get_img_name(), dataset.Dataset::__getDataType(), dataset.Dataset::__getFileInfoList(), dirstructure.Comparison::__make_image(), core.autovars.NTupleVariable::__repr__(), core.autovars.NTupleObjectType::__repr__(), core.autovars.NTupleObject::__repr__(), core.autovars.NTupleCollection::__repr__(), dirstructure.Directory::__repr__(), dqm_interfaces.DirID::__repr__(), dirstructure.Comparison::__repr__(), config.Service::__setattr__(), config.CFG::__str__(), counter.Counter::__str__(), average.Average::__str__(), FWLite.WorkingPoints::_reformat_cut_definitions(), core.autovars.NTupleObjectType::addSubObjects(), core.autovars.NTupleObjectType::addVariables(), core.autovars.NTupleObjectType::allVars(), dataset.CMSDataset::buildListOfFiles(), dataset.LocalDataset::buildListOfFiles(), dataset.CMSDataset::buildListOfFilesDBS(), dirstructure.Directory::calcStats(), validation.Sample::digest(), python.rootplot.utilities.Hist::divide(), python.rootplot.utilities.Hist::divide_wilson(), DisplayManager.DisplayManager::Draw(), core.autovars.NTupleVariable::fillBranch(), core.autovars.NTupleObject::fillBranches(), core.autovars.NTupleCollection::fillBranchesScalar(), core.autovars.NTupleCollection::fillBranchesVector(), core.autovars.NTupleCollection::get_cpp_declaration(), core.autovars.NTupleCollection::get_cpp_wrapper_class(), core.autovars.NTupleCollection::get_py_wrapper_class(), utils.StatisticalTest::get_status(), getName(), production_tasks.Task::getname(), dataset.CMSDataset::getPrimaryDatasetEntries(), dataset.PrivateDataset::getPrimaryDatasetEntries(), VIDSelectorBase.VIDSelectorBase::initialize(), core.autovars.NTupleVariable::makeBranch(), core.autovars.NTupleObject::makeBranches(), core.autovars.NTupleCollection::makeBranchesScalar(), core.autovars.NTupleCollection::makeBranchesVector(), Node(), dirstructure.Directory::print_report(), dataset.BaseDataset::printInfo(), dataset.Dataset::printInfo(), production_tasks.MonitorJobs::run(), setName(), python.rootplot.utilities.Hist::TGraph(), python.rootplot.utilities.Hist::TH1F(), theMiracleOfChildBirth(), counter.Counter::write(), and average.Average::write().

numEvents

int emtf::Node::numEvents

private

Definition at line 79 of file Node.h.

Referenced by calcOptimumSplit(), getNumEvents(), setEvents(), and setNumEvents().

parent

Node* emtf::Node::parent

private

Definition at line 69 of file Node.h.

Referenced by edmIntegrityCheck.PublishToFileSystem::get(), getParent(), Node(), edmIntegrityCheck.PublishToFileSystem::publish(), and setParent().

rightDaughter

Node* emtf::Node::rightDaughter

private

Definition at line 68 of file Node.h.

Referenced by filterEventsToDaughters(), filterEventToDaughter(), getRightDaughter(), Node(), setRightDaughter(), theMiracleOfChildBirth(), and ~Node().

splitValue

double emtf::Node::splitValue

private

Definition at line 71 of file Node.h.

Referenced by calcOptimumSplit(), filterEventsToDaughters(), filterEventToDaughter(), getSplitValue(), Node(), and setSplitValue().

splitVariable

int emtf::Node::splitVariable

private

Definition at line 72 of file Node.h.

Referenced by calcOptimumSplit(), filterEventsToDaughters(), filterEventToDaughter(), getSplitVariable(), Node(), and setSplitVariable().

totalError

double emtf::Node::totalError

private

Definition at line 75 of file Node.h.

Referenced by calcOptimumSplit(), getTotalError(), Node(), and setTotalError().