gum::Estimator< GUM_SCALAR > Class Template Reference

#include <estimator.h>

Collaboration diagram for gum::Estimator< GUM_SCALAR >:

Public Member Functions
	Estimator ()
	Default constructor. More...
	Estimator (const IBayesNet< GUM_SCALAR > *bn)
	Constructor with Bayesian network. More...
	~Estimator ()
GUM_SCALAR	confidence ()
	computes the maximum length of confidence interval for each possible value of each variable More...
void	update (Instantiation I, GUM_SCALAR w)
	updates the estimator with a given sample More...
const Potential< GUM_SCALAR > &	posterior (const DiscreteVariable &var)
	returns the posterior of a node More...
void	clear ()
	refresh the estimator state as empty More...
void	setFromBN (const IBayesNet< GUM_SCALAR > *bn, const NodeSet &hardEvidence)
	estimator initializing More...
void	setFromLBP (LoopyBeliefPropagation< GUM_SCALAR > *lbp, const NodeSet &hardEvidence, GUM_SCALAR virtualLBPSize)
	sets the estimatoor object with posteriors obtained by LoopyBeliefPropagation More...

Detailed Description

template<typename GUM_SCALAR>
class gum::Estimator< GUM_SCALAR >

Definition at line 39 of file estimator.h.

Constructor & Destructor Documentation

Estimator() [1/2]

template<typename GUM_SCALAR >

Estimator::Estimator

(

)

Default constructor.

Definition at line 33 of file estimator_tpl.h.

                                     {
    GUM_CONSTRUCTOR(Estimator);
    wtotal_ = (GUM_SCALAR)0.;
    ntotal_ = (Size)0;
    bn_     = nullptr;
  }

Estimator() [2/2]

template<typename GUM_SCALAR >

Estimator::Estimator ( const IBayesNet< GUM_SCALAR > *  bn )

explicit

Constructor with Bayesian network.

Definition at line 42 of file estimator_tpl.h.

References gum::Set< Key, Alloc >::emplace().

                                                                      : Estimator() {
    bn_ = bn;

    for (gum::NodeGraphPartIterator iter = bn->nodes().begin(); iter != bn->nodes().end(); ++iter)
      estimator_.insert(bn->variable(*iter).name(),
                        std::vector< GUM_SCALAR >(bn->variable(*iter).domainSize(), 0.0));

    GUM_CONSTRUCTOR(Estimator);
  }

Here is the call graph for this function:

~Estimator()

template<typename GUM_SCALAR >

INLINE Estimator::~Estimator

(

)

Definition at line 54 of file estimator_tpl.h.

References gum::Set< Key, Alloc >::emplace().

                                             {
    GUM_DESTRUCTOR(Estimator);
    // remove all the posteriors computed
    clear();
  }

Here is the call graph for this function:

Member Function Documentation

clear()

template<typename GUM_SCALAR >

void Estimator::clear

(

)

refresh the estimator state as empty

this function remove all the statistics in order to restart the computations.

Definition at line 174 of file estimator_tpl.h.

References gum::Set< Key, Alloc >::emplace().

                                      {
    estimator_.clear();
    wtotal_ = (GUM_SCALAR)0;
    ntotal_ = Size(0);
    for (const auto& pot: _target_posteriors_)
      delete pot.second;
    _target_posteriors_.clear();
  }

Here is the call graph for this function:

confidence()

template<typename GUM_SCALAR >

GUM_SCALAR Estimator::confidence

(

)

computes the maximum length of confidence interval for each possible value of each variable

Returns

maximum length of confidence interval

Definition at line 160 of file estimator_tpl.h.

References gum::Set< Key, Alloc >::emplace().

                                                 {
    GUM_SCALAR ic_max = 0;

    for (auto iter = estimator_.begin(); iter != estimator_.end(); ++iter) {
      for (Idx i = 0; i < iter.val().size(); i++) {
        GUM_SCALAR ic = GUM_SCALAR(2 * 1.96 * std::sqrt(variance(iter.key(), i) / (ntotal_ - 1)));
        if (ic > ic_max) ic_max = ic;
      }
    }

    return ic_max;
  }

Here is the call graph for this function:

EV()

template<typename GUM_SCALAR >

GUM_SCALAR Estimator::EV ( std::string  name, Idx  val  )

private

returns expected value of Bernouilli variable (called by it's name) of given parameter

Returns

Expected value of Bernouilli variable (called by it's name) of given parameter

Parameters

name	variable's name, considered as a Bernouilli variable
val	the parameter of the Bernouilli variable

computes the amount of cumulative weights for paramater val over the amount of total cumulative weights

Definition at line 142 of file estimator_tpl.h.

References gum::Set< Key, Alloc >::emplace().

                                                                {
    return estimator_[name][val] / wtotal_;
  }

Here is the call graph for this function:

posterior()

template<typename GUM_SCALAR >

const Potential< GUM_SCALAR > & Estimator::posterior

(

const DiscreteVariable &

var

)

returns the posterior of a node

Returns

a constant ref to the posterior probability of the variable node

Parameters

var	the variable node which we want posterior for

returns the vector of cumulated weight bias for each value of the variable normalized as a CPT

Exceptions

NotFound

if variable node is not in estimator.

Definition at line 119 of file estimator_tpl.h.

References gum::Set< Key, Alloc >::emplace().

                                                                                               {
    Potential< GUM_SCALAR >* p = nullptr;

    if (!estimator_.exists(var.name())) GUM_ERROR(NotFound, "Target variable not found")

    // check if we have already computed the posterior
    if (_target_posteriors_.exists(var.name())) {
      p = _target_posteriors_[var.name()];
    } else {
      p = new Potential< GUM_SCALAR >();
      *p << var;
      _target_posteriors_.insert(var.name(), p);
    }

    p->fillWith(estimator_[var.name()]);
    p->normalize();
    return *p;
  }

Here is the call graph for this function:

setFromBN()

template<typename GUM_SCALAR >

void Estimator::setFromBN	(	const IBayesNet< GUM_SCALAR > *	bn,
		const NodeSet &	hardEvidence
	)

estimator initializing

sets the estimator object with 0-filled vectors corresponding to each non evidence node

Definition at line 64 of file estimator_tpl.h.

References gum::Set< Key, Alloc >::emplace().

                                                                                       {
    for (gum::NodeGraphPartIterator iter = bn->nodes().begin(); iter != bn->nodes().end(); ++iter) {
      auto v = bn->variable(*iter).name();

      if (!hardEvidence.contains(*iter)) {
        if (estimator_.exists(v))
          estimator_[v]
             = std::vector< GUM_SCALAR >(bn->variable(*iter).domainSize(), (GUM_SCALAR)0.0);
        else
          estimator_.insert(
             v,
             std::vector< GUM_SCALAR >(bn->variable(*iter).domainSize(), (GUM_SCALAR)0.0));
      }
    }
  }

Here is the call graph for this function:

setFromLBP()

template<typename GUM_SCALAR >

void Estimator::setFromLBP	(	LoopyBeliefPropagation< GUM_SCALAR > *	lbp,
		const NodeSet &	hardEvidence,
		GUM_SCALAR	virtualLBPSize
	)

sets the estimatoor object with posteriors obtained by LoopyBeliefPropagation

Definition at line 84 of file estimator_tpl.h.

References gum::Set< Key, Alloc >::emplace().

                                                                                                 {
    for (const auto& node: lbp->BN().nodes()) {
      if (!hardEvidence.contains(node)) {
        std::vector< GUM_SCALAR > v;
        auto                      p = lbp->posterior(node);
        gum::Instantiation        inst(p);

        for (inst.setFirst(); !inst.end(); ++inst) {
          v.push_back(p[inst] * virtualLBPSize);
        }

        estimator_.insert(lbp->BN().variable(node).name(), v);
      }
    }
    ntotal_ = (Size)virtualLBPSize;
    wtotal_ = virtualLBPSize;
  }

Here is the call graph for this function:

update()

template<typename GUM_SCALAR >

void Estimator::update	(	Instantiation	I,
		GUM_SCALAR	w
	)

updates the estimator with a given sample

Parameters

I	the sample used to update the estimators
weight	bias for the given sample

adds the sample weight to each node's given value in the estimator

Definition at line 107 of file estimator_tpl.h.

References gum::Set< Key, Alloc >::emplace().

                                                                    {
    wtotal_ += w;
    ntotal_ += (Size)1;

    for (Idx i = 0; i < I.nbrDim(); i++) {
      if (estimator_.exists(I.variable(i).name())) estimator_[I.variable(i).name()][I.val(i)] += w;
    }
  }

Here is the call graph for this function:

variance()

template<typename GUM_SCALAR >

GUM_SCALAR Estimator::variance ( std::string  name, Idx  val  )

private

returns variance of Bernouilli variable (called by it's name) of given parameter

Returns

variance of Bernouilli variable (called by it's name) of given parameter

Parameters

name	variable's name, considered as a Bernouilli variable
val	the parameter of the Bernouilli variable

computes variance for Bernouilli law using EV(name, val)

Definition at line 150 of file estimator_tpl.h.

References gum::Set< Key, Alloc >::emplace().

                                                                      {
    GUM_SCALAR p = EV(name, val);
    return p * (1 - p);
  }

Here is the call graph for this function:

Member Data Documentation

_target_posteriors_

template<typename GUM_SCALAR >

HashTable< std::string, Potential< GUM_SCALAR >* > gum::Estimator< GUM_SCALAR >::_target_posteriors_

private

the set of single posteriors computed during the last inference

the posteriors are owned by LazyPropagation.

Definition at line 157 of file estimator.h.

bn_

template<typename GUM_SCALAR >

const IBayesNet< GUM_SCALAR >* gum::Estimator< GUM_SCALAR >::bn_

private

Bayesian network on which approximation is done.

Definition at line 127 of file estimator.h.

estimator_

template<typename GUM_SCALAR >

HashTable< std::string, std::vector< GUM_SCALAR > > gum::Estimator< GUM_SCALAR >::estimator_

private

estimator represented by hashtable between each variable name and a vector of cumulative sample weights

Definition at line 118 of file estimator.h.

ntotal_

template<typename GUM_SCALAR >

Size gum::Estimator< GUM_SCALAR >::ntotal_

private

number of generated samples

Definition at line 124 of file estimator.h.

wtotal_

template<typename GUM_SCALAR >

GUM_SCALAR gum::Estimator< GUM_SCALAR >::wtotal_

private

cumulated weights of all samples

Definition at line 121 of file estimator.h.

---

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

• agrum/BN/inference/tools/estimator.h
• agrum/BN/inference/tools/estimator_tpl.h