gum::IMarkovRandomField< GUM_SCALAR > Class Template Reference

Class representing the minimal interface for Markov random field. More...

#include <agrum/BN/IMarkovRandomField.h>

Inheritance diagram for gum::IMarkovRandomField< GUM_SCALAR >:

Collaboration diagram for gum::IMarkovRandomField< GUM_SCALAR >:

Public Member Functions
bool	operator== (const IMarkovRandomField< GUM_SCALAR > &from) const
	This operator compares 2 BNs !
bool	operator!= (const IMarkovRandomField< GUM_SCALAR > &from) const
Size	dim () const
	Returns the dimension (the number of free parameters) in this bayes net.
Size	maxVarDomainSize () const
GUM_SCALAR	minParam () const
GUM_SCALAR	maxParam () const
GUM_SCALAR	minNonZeroParam () const
GUM_SCALAR	maxNonOneParam () const
virtual std::string	toDot () const
virtual std::string	toDotAsFactorGraph () const
std::string	toString () const
NodeSet	minimalCondSet (NodeId target, const NodeSet &soids) const
NodeSet	minimalCondSet (const NodeSet &targets, const NodeSet &soids) const
bool	hasSameStructure (const UGmodel &other)
	check if nodes X and nodes Y are independent given nodes Z
NodeSet	family (const NodeId id) const final
	returns the the node as a NodeSet (in mixed graph : family is the node and its parents
NodeSet	family (const std::string &name) const final
double	log10DomainSize () const
Constructors / Destructors
	IMarkovRandomField ()
	Default constructor.
	IMarkovRandomField (std::string name)
	Default constructor.
virtual	~IMarkovRandomField ()
	Destructor.
	IMarkovRandomField (const IMarkovRandomField< GUM_SCALAR > &source)
	Copy constructor.
IMarkovRandomField< GUM_SCALAR > &	operator= (const IMarkovRandomField< GUM_SCALAR > &source)
	Copy operator.
Pure Virtual methods
virtual const Tensor< GUM_SCALAR > &	factor (const NodeSet &varIds) const =0
	Returns the factor of a set of variable.
virtual const NodeSet &	smallestFactorFromNode (NodeId node) const =0
	Returns the smallest factor that contains this variable.
virtual const NodeSet &	smallestFactorFromNode (const std::string &name) const
	Returns the smallest factor that contains this variable.
virtual const FactorTable< GUM_SCALAR > &	factors () const =0
	Returns the set of factors as a IMarkovRandomField::FactorTable.
virtual const VariableNodeMap &	variableNodeMap () const =0
	Returns a constant reference to the VariableNodeMap of thisBN.
virtual const DiscreteVariable &	variable (NodeId id) const =0
	Returns a constant reference over a variable given it's node id.
virtual NodeId	nodeId (const DiscreteVariable &var) const =0
	Return id node from discrete var pointer.
virtual NodeId	idFromName (const std::string &name) const =0
	Getter by name.
virtual const DiscreteVariable &	variableFromName (const std::string &name) const =0
	Getter by name.
Variable manipulation methods.
const UndiGraph &	graph () const
	Returns a constant reference to the dag of this Bayes Net.
virtual Size	size () const final
	Returns the number of variables in this Directed Graphical Model.
Size	sizeEdges () const
	Returns the number of arcs in this Directed Graphical Model.
const NodeGraphPart &	nodes () const final
	Returns a constant reference to the dag of this Bayes Net.
bool	exists (NodeId node) const final
	Return true if this node exists in this graphical model.
bool	exists (const std::string &name) const final
	Returns a constant reference to the dag of this Bayes Net.
Edge manipulation methods.
const EdgeSet &	edges () const
	return true if the edge node1-node2 exists in the UGModel
bool	existsEdge (const NodeId node1, const NodeId node2) const
	return true if the edge node1-node2 exists in the UGModel
bool	existsEdge (const std::string &name1, const std::string &name2) const
	return true if the edge node1-node2 exists in the UGModel
const NodeSet &	neighbours (const NodeId id) const
	returns the neighbours of a node as set of nodes
const NodeSet &	neighbours (const std::string &name) const
	return true if the edge node1-node2 exists in the UGModel
bool	isIndependent (NodeId X, NodeId Y, const NodeSet &Z) const final
	check if X and Y are independent given Z
bool	isIndependent (const NodeSet &X, const NodeSet &Y, const NodeSet &Z) const final
	check if nodes X and nodes Y are independent given nodes Z
bool	isIndependent (const std::string &Xname, const std::string &Yname, const std::vector< std::string > &Znames) const
	return true if the edge node1-node2 exists in the UGModel
bool	isIndependent (const std::vector< std::string > &Xnames, const std::vector< std::string > &Ynames, const std::vector< std::string > &Znames) const
	return true if the edge node1-node2 exists in the UGModel
Getter and setters
const std::string &	property (const std::string &name) const
	Return the value of the property name of this GraphicalModel.
const std::string &	propertyWithDefault (const std::string &name, const std::string &byDefault) const
	Return the value of the property name of this GraphicalModel.
void	setProperty (const std::string &name, const std::string &value)
	Add or change a property of this GraphicalModel.
std::vector< std::string >	properties () const
	List of all the properties.
Variable manipulation methods.
virtual bool	empty () const
	Return true if this graphical model is empty.
std::vector< std::string >	names (const std::vector< NodeId > &ids) const
	transform a vector of NodeId in a vector of names
std::vector< std::string >	names (const NodeSet &ids) const
	transform a NodeSet in a vector of names
std::vector< NodeId >	ids (const std::vector< std::string > &names) const
	transform a vector of names into a vector of nodeId
NodeSet	nodeset (const std::vector< std::string > &names) const
	transform a vector of names into a NodeSet
gum::VariableSet	variables (const std::vector< std::string > &l) const
	transform a vector of names into a VariableeSet
gum::VariableSet	variables (const NodeSet &ids) const
	transform a vector of NodeId into a VariableeSet
Instantiation	completeInstantiation () const
	Get an instantiation over all the variables of the model.

Static Public Member Functions
static void	spaceCplxToStream (std::stringstream &s, double dSize, int dim, Size usedMem)
	send to the stream the space complexity with 3 parametrs

Protected Attributes
UndiGraph	graph_
	The DAG of this Directed Graphical Model.

Private Member Functions
void	_minimalCondSetVisit_ (NodeId node, const NodeSet &soids, NodeSet &minimal, NodeSet &alreadyVisited) const
const HashTable< std::string, std::string > &	_properties_ () const
	Return the properties of this Directed Graphical Model.

Private Attributes
HashTable< std::string, std::string >	_propertiesMap_
	The properties of this Directed Graphical Model.

Detailed Description

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

Class representing the minimal interface for Markov random field.

This class is used as a base class for different versions of Markov random fields. No data (except the graph herited from UGModel are included in this class. Many algorithms (inference for instance) may use this class when a simple MRF is needed.

Definition at line 75 of file IMarkovRandomField.h.

Constructor & Destructor Documentation

IMarkovRandomField() [1/3]

template<typename GUM_SCALAR>

INLINE gum::IMarkovRandomField< GUM_SCALAR >::IMarkovRandomField

(

)

Default constructor.

Definition at line 62 of file IMarkovRandomField_tpl.h.

                                                              : UGmodel() {
    GUM_CONSTRUCTOR(IMarkovRandomField);
  }

References IMarkovRandomField(), and gum::UGmodel::UGmodel().

Referenced by IMarkovRandomField(), IMarkovRandomField(), IMarkovRandomField(), ~IMarkovRandomField(), operator!=(), operator=(), operator==(), and variableFromName().

Here is the call graph for this function:

Here is the caller graph for this function:

IMarkovRandomField() [2/3]

template<typename GUM_SCALAR>

INLINE gum::IMarkovRandomField< GUM_SCALAR >::IMarkovRandomField ( std::string name )

explicit

Default constructor.

Definition at line 67 of file IMarkovRandomField_tpl.h.

                                                                            : UGmodel() {
    GUM_CONSTRUCTOR(IMarkovRandomField);
    this->setProperty("name", name);
  }

References IMarkovRandomField(), gum::UGmodel::UGmodel(), and gum::GraphicalModel::setProperty().

Here is the call graph for this function:

~IMarkovRandomField()

template<typename GUM_SCALAR>

gum::IMarkovRandomField< GUM_SCALAR >::~IMarkovRandomField ( )

virtual

Destructor.

Definition at line 87 of file IMarkovRandomField_tpl.h.

                                                        {
    GUM_DESTRUCTOR(IMarkovRandomField);
  }

References IMarkovRandomField().

Here is the call graph for this function:

IMarkovRandomField() [3/3]

template<typename GUM_SCALAR>

gum::IMarkovRandomField< GUM_SCALAR >::IMarkovRandomField

(

const IMarkovRandomField< GUM_SCALAR > &

source

)

Copy constructor.

Definition at line 73 of file IMarkovRandomField_tpl.h.

                                                      : UGmodel(source) {
    GUM_CONS_CPY(IMarkovRandomField);
  }

References IMarkovRandomField(), and gum::UGmodel::UGmodel().

Here is the call graph for this function:

Member Function Documentation

_minimalCondSetVisit_()

template<typename GUM_SCALAR>

void gum::IMarkovRandomField< GUM_SCALAR >::_minimalCondSetVisit_ ( NodeId node, const NodeSet & soids, NodeSet & minimal, NodeSet & alreadyVisited ) const

private

Definition at line 336 of file IMarkovRandomField_tpl.h.

                                                                                              {
    if (alreadyVisited.contains(node)) return;
    alreadyVisited << node;
 
    if (soids.contains(node)) {
      minimal << node;
    } else {
      for (auto neig: graph_.neighbours(node))
        _minimalCondSetVisit_(neig, soids, minimal, alreadyVisited);
    }
  }

References _minimalCondSetVisit_(), gum::Set< Key >::contains(), and gum::UGmodel::graph_.

Referenced by _minimalCondSetVisit_(), and minimalCondSet().

Here is the call graph for this function:

Here is the caller graph for this function:

_properties_()

INLINE const HashTable< std::string, std::string > & gum::GraphicalModel::_properties_ ( ) const

privateinherited

Return the properties of this Directed Graphical Model.

Definition at line 70 of file graphicalModel_inl.h.

                                                                              {
    return _propertiesMap_;
  }

References _propertiesMap_.

Referenced by property().

Here is the caller graph for this function:

completeInstantiation()

INLINE Instantiation gum::GraphicalModel::completeInstantiation ( ) const

inherited

Get an instantiation over all the variables of the model.

Definition at line 106 of file graphicalModel_inl.h.

                                                            {
    Instantiation I;
 
    for (const auto node: nodes())
      I << variable(node);
 
    return I;
  }

References nodes().

Here is the call graph for this function:

dim()

template<typename GUM_SCALAR>

INLINE Size gum::IMarkovRandomField< GUM_SCALAR >::dim

(

)

const

Returns the dimension (the number of free parameters) in this bayes net.

\( dim(G)=\sum_{i \in nodes} ((r_i-1)\cdot q_i) \) where \( r_i \) is the number of instantiations of node \( i \) and \( q_i \) is the number of instantiations of its parents.

Definition at line 92 of file IMarkovRandomField_tpl.h.

                                                          {
    Size res = 0;
    for (auto f: factors()) {
      res += f.second->domainSize();
    }
    return res;
  }

References factors().

Here is the call graph for this function:

edges()

INLINE const EdgeSet & gum::UGmodel::edges ( ) const

inherited

return true if the edge node1-node2 exists in the UGModel

Parameters

node1	the nodeId (or the name) of the node1
node2	the nodeId (or the name) of the node2

Returns

true if the edge exists

Definition at line 65 of file UGmodel_inl.h.

{ return graph_.edges(); }

References graph_.

Referenced by hasSameStructure().

Here is the caller graph for this function:

empty()

INLINE bool gum::GraphicalModel::empty ( ) const

virtualinherited

Return true if this graphical model is empty.

Definition at line 116 of file graphicalModel_inl.h.

{ return size() == 0; }

References size().

Referenced by gum::IBayesNet< GUM_SCALAR >::check(), and gum::BayesNet< GUM_SCALAR >::clear().

Here is the call graph for this function:

Here is the caller graph for this function:

exists() [1/2]

INLINE bool gum::UGmodel::exists ( const std::string & name ) const

finalvirtualinherited

Returns a constant reference to the dag of this Bayes Net.

Implements gum::GraphicalModel.

Definition at line 83 of file UGmodel_inl.h.

                                                         {
    try {
      return exists(idFromName(name));
    } catch ([[maybe_unused]] gum::NotFound& e) { return false; }
  }

References exists(), and gum::GraphicalModel::idFromName().

Here is the call graph for this function:

exists() [2/2]

INLINE bool gum::UGmodel::exists ( NodeId node ) const

finalvirtualinherited

Return true if this node exists in this graphical model.

Implements gum::GraphicalModel.

Definition at line 81 of file UGmodel_inl.h.

{ return graph_.exists(node); }

References graph_.

Referenced by exists().

Here is the caller graph for this function:

existsEdge() [1/2]

INLINE bool gum::UGmodel::existsEdge ( const NodeId node1, const NodeId node2 ) const

inherited

return true if the edge node1-node2 exists in the UGModel

Parameters

node1	the nodeId (or the name) of the node1
node2	the nodeId (or the name) of the node2

Returns

true if the edge exists

Definition at line 67 of file UGmodel_inl.h.

                                                                              {
    return graph_.existsEdge(node1, node2);
  }

References graph_.

Referenced by existsEdge().

Here is the caller graph for this function:

existsEdge() [2/2]

INLINE bool gum::UGmodel::existsEdge ( const std::string & name1, const std::string & name2 ) const

inherited

return true if the edge node1-node2 exists in the UGModel

Parameters

node1	the nodeId (or the name) of the node1
node2	the nodeId (or the name) of the node2

Returns

true if the edge exists

Definition at line 71 of file UGmodel_inl.h.

                                                                                      {
    return existsEdge(idFromName(name1), idFromName(name2));
  }

References existsEdge(), and gum::GraphicalModel::idFromName().

Here is the call graph for this function:

factor()

template<typename GUM_SCALAR>

virtual const Tensor< GUM_SCALAR > & gum::IMarkovRandomField< GUM_SCALAR >::factor ( const NodeSet & varIds ) const

pure virtual

Returns the factor of a set of variable.

Exceptions

NotFound

If no variable's id matches varId.

Referenced by operator==(), and toString().

Here is the caller graph for this function:

factors()

template<typename GUM_SCALAR>

virtual const FactorTable< GUM_SCALAR > & gum::IMarkovRandomField< GUM_SCALAR >::factors ( ) const

pure virtual

Returns the set of factors as a IMarkovRandomField::FactorTable.

Referenced by dim(), maxNonOneParam(), maxParam(), minNonZeroParam(), minParam(), operator==(), toDotAsFactorGraph(), and toString().

Here is the caller graph for this function:

family() [1/2]

INLINE NodeSet gum::UGmodel::family ( const NodeId id ) const

finalvirtualinherited

returns the the node as a NodeSet (in mixed graph : family is the node and its parents

Parameters

id	the node which is the head of an arc with the returned nodes
name	the name of the node the node which is the head of an arc with the returned nodes

Implements gum::GraphicalModel.

Definition at line 99 of file UGmodel_inl.h.

                                                      {
    // family is the set of the note and its parent. So for a completely undirected model, it is
    // just the node.
    return gum::NodeSet{id};
  }

Referenced by family().

Here is the caller graph for this function:

family() [2/2]

INLINE NodeSet gum::UGmodel::family ( const std::string & name ) const

finalvirtualinherited

Implements gum::GraphicalModel.

Definition at line 105 of file UGmodel_inl.h.

{ return family(idFromName(name)); }

References family(), and gum::GraphicalModel::idFromName().

Here is the call graph for this function:

graph()

INLINE const UndiGraph & gum::UGmodel::graph ( ) const

inherited

Returns a constant reference to the dag of this Bayes Net.

Definition at line 57 of file UGmodel_inl.h.

{ return graph_; }

References graph_.

Referenced by gum::MarginalTargetedMRFInference< GUM_SCALAR >::MarginalTargetedMRFInference(), isIndependent(), isIndependent(), size(), and gum::IMarkovRandomField< GUM_SCALAR >::toString().

Here is the caller graph for this function:

hasSameStructure()

bool gum::UGmodel::hasSameStructure ( const UGmodel & other )

inherited

check if nodes X and nodes Y are independent given nodes Z

Returns

true if all the named node are the same and all the named arcs are the same

Definition at line 70 of file UGmodel.cpp.

                                                     {
    if (this == &other) return true;
 
    if (size() != other.size()) return false;
 
    if (sizeEdges() != other.sizeEdges()) return false;
 
    for (const auto& nid: nodes()) {
      try {
        other.idFromName(variable(nid).name());
      } catch (NotFound const&) { return false; }
    }
 
    for (const auto& edge: edges()) {
      if (!other.edges().exists(Edge(other.idFromName(variable(edge.first()).name()),
                                     other.idFromName(variable(edge.second()).name()))))
        return false;
    }
 
    return true;
  }

References UGmodel(), edges(), gum::Set< Key >::exists(), gum::GraphicalModel::idFromName(), nodes(), size(), sizeEdges(), and gum::GraphicalModel::variable().

Here is the call graph for this function:

idFromName()

template<typename GUM_SCALAR>

virtual NodeId gum::IMarkovRandomField< GUM_SCALAR >::idFromName ( const std::string & name ) const

pure virtual

Getter by name.

Exceptions

NotFound

if no such name exists in the graph.

Implements gum::GraphicalModel.

Referenced by operator==(), and smallestFactorFromNode().

Here is the caller graph for this function:

ids()

INLINE std::vector< NodeId > gum::GraphicalModel::ids ( const std::vector< std::string > & names ) const

inherited

transform a vector of names into a vector of nodeId

Returns

the vector of names

Definition at line 139 of file graphicalModel_inl.h.

                                                                                   {
    std::vector< NodeId >  res;
    const VariableNodeMap& v = variableNodeMap();
    std::transform(names.cbegin(),
                   names.cend(),
                   std::back_inserter(res),
                   [v](const std::string& n) { return v.idFromName(n); });
    return res;
  }

References names().

Referenced by gum::DAGmodel::children(), exists(), names(), and gum::DAGmodel::parents().

Here is the call graph for this function:

Here is the caller graph for this function:

isIndependent() [1/4]

INLINE bool gum::UGmodel::isIndependent ( const NodeSet & X, const NodeSet & Y, const NodeSet & Z ) const

finalvirtualinherited

check if nodes X and nodes Y are independent given nodes Z

Implements gum::GraphicalModel.

Definition at line 95 of file UGmodel_inl.h.

                                                                                               {
    return !graph().hasUndirectedPath(X, Y, Z);
  }

References graph(), and gum::EdgeGraphPart::hasUndirectedPath().

Here is the call graph for this function:

isIndependent() [2/4]

bool gum::UGmodel::isIndependent ( const std::string & Xname, const std::string & Yname, const std::vector< std::string > & Znames ) const

inlineinherited

return true if the edge node1-node2 exists in the UGModel

Parameters

node1	the nodeId (or the name) of the node1
node2	the nodeId (or the name) of the node2

Returns

true if the edge exists

Definition at line 144 of file UGmodel.h.

                                                                   {
      return isIndependent(idFromName(Xname), idFromName(Yname), nodeset(Znames));
    };

References gum::GraphicalModel::idFromName(), isIndependent(), and gum::GraphicalModel::nodeset().

Here is the call graph for this function:

isIndependent() [3/4]

bool gum::UGmodel::isIndependent ( const std::vector< std::string > & Xnames, const std::vector< std::string > & Ynames, const std::vector< std::string > & Znames ) const

inlineinherited

return true if the edge node1-node2 exists in the UGModel

Parameters

node1	the nodeId (or the name) of the node1
node2	the nodeId (or the name) of the node2

Returns

true if the edge exists

Definition at line 150 of file UGmodel.h.

                                                                   {
      return isIndependent(nodeset(Xnames), nodeset(Ynames), nodeset(Znames));
    };

References isIndependent(), and gum::GraphicalModel::nodeset().

Here is the call graph for this function:

isIndependent() [4/4]

INLINE bool gum::UGmodel::isIndependent ( NodeId X, NodeId Y, const NodeSet & Z ) const

finalvirtualinherited

check if X and Y are independent given Z

Implements gum::GraphicalModel.

Definition at line 91 of file UGmodel_inl.h.

                                                                               {
    return !graph().hasUndirectedPath(X, Y, Z);
  }

References graph(), and gum::EdgeGraphPart::hasUndirectedPath().

Referenced by isIndependent(), and isIndependent().

Here is the call graph for this function:

Here is the caller graph for this function:

log10DomainSize()

INLINE double gum::GraphicalModel::log10DomainSize ( ) const

inherited

Definition at line 95 of file graphicalModel_inl.h.

                                               {
    double dSize = 0.0;
 
    for (const auto node: nodes()) {
      dSize += std::log10(variable(node).domainSize());
    }
 
    return dSize;
  }

References nodes().

Referenced by gum::IBayesNet< GUM_SCALAR >::toString(), gum::IMarkovRandomField< GUM_SCALAR >::toString(), and gum::InfluenceDiagram< GUM_SCALAR >::toString().

Here is the call graph for this function:

Here is the caller graph for this function:

maxNonOneParam()

template<typename GUM_SCALAR>

INLINE GUM_SCALAR gum::IMarkovRandomField< GUM_SCALAR >::maxNonOneParam

(

)

const

Returns

the biggest value (not equal to 1) in the CPTs of *this

Definition at line 141 of file IMarkovRandomField_tpl.h.

                                                                           {
    GUM_SCALAR res = 0.0;
    for (auto elt: factors()) {
      auto v = elt.second->maxNonOne();
      if (v > res) { res = v; }
    }
    return res;
  }

References factors().

Here is the call graph for this function:

maxParam()

template<typename GUM_SCALAR>

INLINE GUM_SCALAR gum::IMarkovRandomField< GUM_SCALAR >::maxParam

(

)

const

Returns

the biggest value in the CPTs of *this

Definition at line 121 of file IMarkovRandomField_tpl.h.

                                                                     {
    GUM_SCALAR res = 1.0;
    for (auto elt: factors()) {
      auto v = elt.second->max();
      if (v > res) { res = v; }
    }
    return res;
  }

References factors().

Here is the call graph for this function:

maxVarDomainSize()

template<typename GUM_SCALAR>

INLINE Size gum::IMarkovRandomField< GUM_SCALAR >::maxVarDomainSize

(

)

const

Returns

the biggest domainSize among the variables of *this

Definition at line 101 of file IMarkovRandomField_tpl.h.

                                                                       {
    Size res = 0;
    for (auto node: nodes()) {
      auto v = variable(node).domainSize();
      if (v > res) { res = v; }
    }
    return res;
  }

References gum::UGmodel::nodes(), and variable().

Here is the call graph for this function:

minimalCondSet() [1/2]

template<typename GUM_SCALAR>

NodeSet gum::IMarkovRandomField< GUM_SCALAR >::minimalCondSet	(	const NodeSet &	targets,
		const NodeSet &	soids ) const

Definition at line 366 of file IMarkovRandomField_tpl.h.

                                                                                       {
    NodeSet res;
    for (auto node: targets) {
      res += minimalCondSet(node, soids);
    }
    return res;
  }

References minimalCondSet().

Here is the call graph for this function:

minimalCondSet() [2/2]

template<typename GUM_SCALAR>

NodeSet gum::IMarkovRandomField< GUM_SCALAR >::minimalCondSet	(	NodeId	target,
		const NodeSet &	soids ) const

Definition at line 352 of file IMarkovRandomField_tpl.h.

                                                                                       {
    if (soids.contains(target)) return NodeSet({target});
 
    NodeSet res;
    NodeSet alreadyVisited;
    alreadyVisited << target;
 
    for (auto neig: graph_.neighbours(target))
      _minimalCondSetVisit_(neig, soids, res, alreadyVisited);
    return res;
  }

References _minimalCondSetVisit_(), gum::Set< Key >::contains(), and gum::UGmodel::graph_.

Referenced by minimalCondSet().

Here is the call graph for this function:

Here is the caller graph for this function:

minNonZeroParam()

template<typename GUM_SCALAR>

INLINE GUM_SCALAR gum::IMarkovRandomField< GUM_SCALAR >::minNonZeroParam

(

)

const

Returns

the smallest value (not equal to 0) in the CPTs of *this

Definition at line 131 of file IMarkovRandomField_tpl.h.

                                                                            {
    GUM_SCALAR res = 1.0;
    for (auto elt: factors()) {
      auto v = elt.second->minNonZero();
      if (v < res) { res = v; }
    }
    return res;
  }

References factors().

Here is the call graph for this function:

minParam()

template<typename GUM_SCALAR>

INLINE GUM_SCALAR gum::IMarkovRandomField< GUM_SCALAR >::minParam

(

)

const

Returns

the smallest value in the CPTs of *this

Definition at line 111 of file IMarkovRandomField_tpl.h.

                                                                     {
    GUM_SCALAR res = 1.0;
    for (auto elt: factors()) {
      auto v = elt.second->min();
      if (v < res) { res = v; }
    }
    return res;
  }

References factors().

Here is the call graph for this function:

names() [1/2]

INLINE std::vector< std::string > gum::GraphicalModel::names ( const NodeSet & ids ) const

inherited

transform a NodeSet in a vector of names

Returns

the vector of names

Definition at line 129 of file graphicalModel_inl.h.

                                                                       {
    const VariableNodeMap&     v = variableNodeMap();
    std::vector< std::string > res;
    for (auto n: ids) {
      res.push_back(v.name(n));
    }
    return res;
  }

names() [2/2]

INLINE std::vector< std::string > gum::GraphicalModel::names ( const std::vector< NodeId > & ids ) const

inherited

transform a vector of NodeId in a vector of names

Returns

the vector of names

Definition at line 119 of file graphicalModel_inl.h.

                                                                                   {
    std::vector< std::string > res;
    const VariableNodeMap&     v = variableNodeMap();
    std::transform(ids.cbegin(), ids.cend(), std::back_inserter(res), [v](NodeId n) {
      return v[n].name();
    });
    return res;
  }

References ids(), and variableNodeMap().

Referenced by gum::DAGmodel::children(), exists(), ids(), nodeset(), and gum::DAGmodel::parents().

Here is the call graph for this function:

Here is the caller graph for this function:

neighbours() [1/2]

INLINE const NodeSet & gum::UGmodel::neighbours ( const NodeId id ) const

inherited

returns the neighbours of a node as set of nodes

Note that the set of nodes returned may be empty if no edge within the EdgeGraphPart contains the given node.

Parameters

id	the node toward which the edge returned are pointing

Definition at line 75 of file UGmodel_inl.h.

{ return graph_.neighbours(id); }

References graph_.

Referenced by neighbours(), and gum::IMarkovRandomField< GUM_SCALAR >::toDot().

Here is the caller graph for this function:

neighbours() [2/2]

INLINE const NodeSet & gum::UGmodel::neighbours ( const std::string & name ) const

inherited

return true if the edge node1-node2 exists in the UGModel

Parameters

node1	the nodeId (or the name) of the node1
node2	the nodeId (or the name) of the node2

Returns

true if the edge exists

Definition at line 77 of file UGmodel_inl.h.

                                                                       {
    return neighbours(idFromName(name));
  }

References gum::GraphicalModel::idFromName(), and neighbours().

Here is the call graph for this function:

nodeId()

template<typename GUM_SCALAR>

virtual NodeId gum::IMarkovRandomField< GUM_SCALAR >::nodeId ( const DiscreteVariable & var ) const

pure virtual

Return id node from discrete var pointer.

Exceptions

NotFound

If no variable matches var.

Implements gum::GraphicalModel.

nodes()

INLINE const NodeGraphPart & gum::UGmodel::nodes ( ) const

finalvirtualinherited

Returns a constant reference to the dag of this Bayes Net.

Implements gum::GraphicalModel.

Definition at line 89 of file UGmodel_inl.h.

{ return (NodeGraphPart&)graph_; }

References graph_.

Referenced by hasSameStructure(), gum::IMarkovRandomField< GUM_SCALAR >::maxVarDomainSize(), gum::IMarkovRandomField< GUM_SCALAR >::operator==(), gum::IMarkovRandomField< GUM_SCALAR >::toDot(), and gum::IMarkovRandomField< GUM_SCALAR >::toDotAsFactorGraph().

Here is the caller graph for this function:

nodeset()

NodeSet gum::GraphicalModel::nodeset ( const std::vector< std::string > & names ) const

inherited

transform a vector of names into a NodeSet

Returns

NodeSet

Definition at line 60 of file graphicalModel.cpp.

                                                                           {
    NodeSet res;
    for (const auto& name: names) {
      res.insert(idFromName(name));
    }
    return res;
  }

References idFromName(), gum::Set< Key >::insert(), and names().

Referenced by gum::IBayesNet< double >::children(), gum::IBayesNet< double >::children(), gum::DAGmodel::isIndependent(), gum::DAGmodel::isIndependent(), gum::UGmodel::isIndependent(), gum::UGmodel::isIndependent(), gum::DAGmodel::moralizedAncestralGraph(), and gum::DAGmodel::parents().

Here is the call graph for this function:

Here is the caller graph for this function:

operator!=()

template<typename GUM_SCALAR>

INLINE bool gum::IMarkovRandomField< GUM_SCALAR >::operator!=

(

const IMarkovRandomField< GUM_SCALAR > &

from

)

const

Returns

Returns false if the src and this are equal.

Definition at line 313 of file IMarkovRandomField_tpl.h.

                                                                                             {
    return !this->operator==(from);
  }

References IMarkovRandomField(), and gum::operator==().

Here is the call graph for this function:

operator=()

template<typename GUM_SCALAR>

IMarkovRandomField< GUM_SCALAR > & gum::IMarkovRandomField< GUM_SCALAR >::operator=

(

const IMarkovRandomField< GUM_SCALAR > &

source

)

Copy operator.

Definition at line 80 of file IMarkovRandomField_tpl.h.

                                                                                              {
    if (this != &source) { UGmodel::operator=(source); }
 
    return *this;
  }

References IMarkovRandomField(), and gum::UGmodel::operator=().

Here is the call graph for this function:

operator==()

template<typename GUM_SCALAR>

bool gum::IMarkovRandomField< GUM_SCALAR >::operator==

(

const IMarkovRandomField< GUM_SCALAR > &

from

)

const

This operator compares 2 BNs !

Warning

To identify nodes between BNs, it is assumed that they share the same name.

Returns

true if the src and this are equal.

Definition at line 263 of file IMarkovRandomField_tpl.h.

                                                                                      {
    if (size() != from.size()) { return false; }
 
    if (sizeEdges() != from.sizeEdges()) { return false; }
 
    // alignment of variables between the 2 BNs
    Bijection< const DiscreteVariable*, const DiscreteVariable* > alignment;
 
    for (auto node: nodes()) {
      try {
        const auto& v1 = variable(node);
        const auto& v2 = from.variableFromName(variable(node).name());
        if (v1 != v2) return false;
 
        alignment.insert(&variable(node), &from.variableFromName(variable(node).name()));
      } catch (NotFound const&) {
        // a name is not found in from
        return false;
      }
    }
 
    for (const auto& elt: factors()) {
      const auto& key    = elt.first;
      const auto& factor = *elt.second;
 
      NodeSet fromkey;
      for (const auto n: key)
        fromkey.insert(from.idFromName(variable(n).name()));
 
      if (!from.factors().exists(fromkey)) { return false; }
 
      const auto& fromfactor = from.factor(fromkey);
 
      Instantiation i(factor);
      Instantiation j(fromfactor);
      for (i.setFirst(); !i.end(); i.inc()) {
        for (Idx indice = 0; indice < factor.nbrDim(); ++indice) {
          const DiscreteVariable* p = &(i.variable(indice));
          j.chgVal(*(alignment.second(p)), i.val(*p));
        }
 
        if (std::pow(factor.get(i) - fromfactor.get(j), (GUM_SCALAR)2) > (GUM_SCALAR)1e-6) {
          return false;
        }
      }
    }
    return true;
  }

References IMarkovRandomField(), gum::Instantiation::chgVal(), gum::Instantiation::end(), factor(), factors(), idFromName(), gum::Instantiation::inc(), gum::BijectionImplementation< T1, T2, Gen >::insert(), gum::Set< Key >::insert(), gum::UGmodel::nodes(), gum::BijectionImplementation< T1, T2, Gen >::second(), gum::Instantiation::setFirst(), gum::UGmodel::size(), gum::UGmodel::sizeEdges(), gum::Instantiation::val(), variable(), gum::Instantiation::variable(), and variableFromName().

Here is the call graph for this function:

properties()

INLINE std::vector< std::string > gum::GraphicalModel::properties ( ) const

inherited

List of all the properties.

Definition at line 81 of file graphicalModel_inl.h.

                                                          {
    std::vector< std::string > prop;
    for (const auto& [p, v]: _propertiesMap_)
      prop.push_back(p);
    return prop;
  }

References _propertiesMap_.

property()

INLINE const std::string & gum::GraphicalModel::property ( const std::string & name ) const

inherited

Return the value of the property name of this GraphicalModel.

Exceptions

NotFound

Raised if no name property is found.

Definition at line 60 of file graphicalModel_inl.h.

                                                                       {
    try {
      return _properties_()[name];
    } catch (NotFound const&) {
      std::string msg = "The following property does not exists: ";
      GUM_ERROR(NotFound, msg + name)
    }
  }

References _properties_(), and GUM_ERROR.

Referenced by gum::IMarkovRandomField< GUM_SCALAR >::toDot(), gum::InfluenceDiagram< GUM_SCALAR >::toDot(), and gum::IMarkovRandomField< GUM_SCALAR >::toDotAsFactorGraph().

Here is the call graph for this function:

Here is the caller graph for this function:

propertyWithDefault()

INLINE const std::string & gum::GraphicalModel::propertyWithDefault ( const std::string & name, const std::string & byDefault ) const

inherited

Return the value of the property name of this GraphicalModel.

return byDefault if the property name is not found

Definition at line 75 of file graphicalModel_inl.h.

                                                                                           {
    return (_propertiesMap_.exists(name)) ? _propertiesMap_[name] : byDefault;
  }

References _propertiesMap_.

setProperty()

INLINE void gum::GraphicalModel::setProperty ( const std::string & name, const std::string & value )

inherited

Add or change a property of this GraphicalModel.

Definition at line 89 of file graphicalModel_inl.h.

                                                                                {
    if (_propertiesMap_.exists(name)) _propertiesMap_[name] = value;
    else _propertiesMap_.insert(name, value);
  }

References _propertiesMap_.

Referenced by gum::IBayesNet< GUM_SCALAR >::IBayesNet(), gum::IMarkovRandomField< GUM_SCALAR >::IMarkovRandomField(), and gum::InfluenceDiagram< GUM_SCALAR >::fastPrototype().

Here is the caller graph for this function:

size()

INLINE Size gum::UGmodel::size ( ) const

finalvirtualinherited

Returns the number of variables in this Directed Graphical Model.

Implements gum::GraphicalModel.

Definition at line 60 of file UGmodel_inl.h.

{ return graph().size(); }

References graph(), and gum::NodeGraphPart::size().

Referenced by hasSameStructure(), gum::IMarkovRandomField< GUM_SCALAR >::operator==(), gum::IMarkovRandomField< GUM_SCALAR >::toDot(), and gum::IMarkovRandomField< GUM_SCALAR >::toString().

Here is the call graph for this function:

Here is the caller graph for this function:

sizeEdges()

INLINE Size gum::UGmodel::sizeEdges ( ) const

inherited

Returns the number of arcs in this Directed Graphical Model.

Definition at line 63 of file UGmodel_inl.h.

{ return graph_.sizeEdges(); }

References graph_.

Referenced by hasSameStructure(), and gum::IMarkovRandomField< GUM_SCALAR >::operator==().

Here is the caller graph for this function:

smallestFactorFromNode() [1/2]

template<typename GUM_SCALAR>

INLINE const NodeSet & gum::IMarkovRandomField< GUM_SCALAR >::smallestFactorFromNode ( const std::string & name ) const

virtual

Returns the smallest factor that contains this variable.

Exceptions

NotFound

If no variable's id matches varId.

Definition at line 326 of file IMarkovRandomField_tpl.h.

                                                                                          {
    try {
      return smallestFactorFromNode(idFromName(name));
    } catch (NotFound const&) {
      GUM_ERROR(NotFound, "No factor containing the variable <" << name << ">")
    }
  }

References GUM_ERROR, idFromName(), and smallestFactorFromNode().

Here is the call graph for this function:

smallestFactorFromNode() [2/2]

template<typename GUM_SCALAR>

virtual const NodeSet & gum::IMarkovRandomField< GUM_SCALAR >::smallestFactorFromNode ( NodeId node ) const

pure virtual

Returns the smallest factor that contains this variable.

Exceptions

NotFound

If no variable's id matches varId.

Referenced by smallestFactorFromNode().

Here is the caller graph for this function:

spaceCplxToStream()

void gum::GraphicalModel::spaceCplxToStream ( std::stringstream & s, double dSize, int dim, Size usedMem )

staticinherited

send to the stream the space complexity with 3 parametrs

Parameters

s	the stream
dSize	the log10domainSize
dim	the dimension
usedMem	the memory needed for the params

Definition at line 69 of file graphicalModel.cpp.

                                                                                               {
    if (dSize > 6) s << "domainSize: 10^" << dSize;
    else s << "domainSize: " << std::round(std::pow(10.0, dSize));
 
    s << ", dim: " << dim << ", mem: ";
 
    if (const Size go = usedMem / (1024 * 1024 * 1024); go > 0) s << go << "Go ";
    if (const Size mo = (usedMem / (1024 * 1024)) % 1024; mo > 0) s << mo << "Mo ";
    if (const Size ko = (usedMem / 1024) % 1024; ko > 0) s << ko << "Ko ";
    s << usedMem % 1024 << "o";
  }

Referenced by gum::IBayesNet< GUM_SCALAR >::toString().

Here is the caller graph for this function:

toDot()

template<typename GUM_SCALAR>

std::string gum::IMarkovRandomField< GUM_SCALAR >::toDot ( ) const

virtual

Returns

Returns a dot representation of this IMarkovRandomField.

Definition at line 170 of file IMarkovRandomField_tpl.h.

                                                          {
    std::stringstream output;
    output << "graph \"";
 
    std::string mn_name;
 
    try {
      mn_name = this->property("name");
    } catch (NotFound const&) { mn_name = "no_name"; }
 
    output << mn_name << "\" {" << std::endl;
    output << "  graph [bgcolor=transparent,label=\"" << mn_name << "\"];" << std::endl;
    output << "  node [style=filled fillcolor=\"#ffffaa\"];" << std::endl << std::endl;
 
    for (auto node: nodes())
      output << "  \"" << variable(node).name() << "\" [comment=\"" << node << ":"
             << variable(node).toStringWithDescription() << "\"];" << std::endl;
 
    output << std::endl;
 
    std::string tab = "  ";
 
    for (auto node: nodes()) {
      if (neighbours(node).size() > 0) {
        for (auto nei: neighbours(node)) {
          if (variable(node).name() < variable(nei).name()) {
            output << tab << "\"" << variable(node).name() << "\" -- "
                   << "\"" << variable(nei).name() << "\";" << std::endl;
          }
        }
      } else {
        output << tab << "\"" << variable(node).name() << "\";" << std::endl;
      }
    }
 
    output << "}" << std::endl;
 
    return output.str();
  }

References gum::UGmodel::neighbours(), gum::UGmodel::nodes(), gum::GraphicalModel::property(), gum::UGmodel::size(), and variable().

Here is the call graph for this function:

toDotAsFactorGraph()

template<typename GUM_SCALAR>

std::string gum::IMarkovRandomField< GUM_SCALAR >::toDotAsFactorGraph ( ) const

virtual

Returns

Returns a dot representation of this IMarkovRandomField.

Definition at line 211 of file IMarkovRandomField_tpl.h.

                                                                       {
    std::stringstream output;
    std::string       mn_name;
    try {
      mn_name = this->property("name");
    } catch (NotFound const&) { mn_name = "no_name"; }
 
    output << "graph FG_" << mn_name << " {" << std::endl;
    output << "  layout=neato;" << std::endl;
    output << "  graph [bgcolor=transparent,label=\"factor graph for " << mn_name << "\"];"
           << std::endl;
 
    // the variables
    output << "  node [shape=rectangle,margin=0.04,width=0,height=0, "
              "style=filled,color=\"coral\"];"
           << std::endl;
    for (auto nod: nodes()) {
      output << "\"" << variable(nod).name() << "\";" << std::endl;
    }
    output << std::endl;
 
    // the factor
    output << "node[shape = point,width = 0.1,height = 0.1,style = filled,color = "
              "\"burlywood\"];"
           << std::endl;
    for (const auto& kv: factors()) {
      output << "  \"f";
      for (NodeId nod: kv.first) {
        output << "#" << variable(nod).name();
      }
      output << "\";" << std::endl;
    }
 
    // the link variable--factors
    output << "  edge[len = 0.7];" << std::endl;
    for (const auto& kv: factors()) {
      std::string clicname = "\"f";
      for (NodeId nod: kv.first) {
        clicname += "#";
        clicname += variable(nod).name();
      }
      clicname += "\"";
 
      for (NodeId nod: kv.first)
        output << "  " << clicname << " -- \"" << variable(nod).name() << "\";" << std::endl;
    }
    output << "}" << std::endl;
 
    return output.str();
  }

References factors(), gum::UGmodel::nodes(), gum::GraphicalModel::property(), and variable().

Here is the call graph for this function:

toString()

template<typename GUM_SCALAR>

INLINE std::string gum::IMarkovRandomField< GUM_SCALAR >::toString

(

)

const

Returns

Returns a string representation of this IMarkovRandomField.

Definition at line 151 of file IMarkovRandomField_tpl.h.

                                                                    {
    Size   param = 0;
    double dSize = log10DomainSize();
 
    for (auto factor: factors())
      param += factor.second->content()->realSize();
 
    std::stringstream s;
    s << "MRF{nodes: " << size() << ", edges: " << graph().sizeEdges() << ", ";
 
    if (dSize > 6) s << "domainSize: 10^" << dSize;
    else s << "domainSize: " << std::round(std::pow(10.0, dSize));
 
    s << ", dim: " << param << "}";
 
    return s.str();
  }

References factor(), factors(), gum::UGmodel::graph(), gum::GraphicalModel::log10DomainSize(), and gum::UGmodel::size().

Referenced by gum::operator<<().

Here is the call graph for this function:

Here is the caller graph for this function:

variable()

template<typename GUM_SCALAR>

virtual const DiscreteVariable & gum::IMarkovRandomField< GUM_SCALAR >::variable ( NodeId id ) const

pure virtual

Returns a constant reference over a variable given it's node id.

Exceptions

NotFound

If no variable's id matches varId.

Implements gum::GraphicalModel.

Referenced by maxVarDomainSize(), operator==(), toDot(), and toDotAsFactorGraph().

Here is the caller graph for this function:

variableFromName()

template<typename GUM_SCALAR>

virtual const DiscreteVariable & gum::IMarkovRandomField< GUM_SCALAR >::variableFromName ( const std::string & name ) const

pure virtual

Getter by name.

Exceptions

NotFound

if no such name exists in the graph.

Implements gum::GraphicalModel.

References IMarkovRandomField().

Referenced by operator==().

Here is the call graph for this function:

Here is the caller graph for this function:

variableNodeMap()

template<typename GUM_SCALAR>

virtual const VariableNodeMap & gum::IMarkovRandomField< GUM_SCALAR >::variableNodeMap ( ) const

pure virtual

Returns a constant reference to the VariableNodeMap of thisBN.

Implements gum::GraphicalModel.

variables() [1/2]

INLINE VariableSet gum::GraphicalModel::variables ( const NodeSet & ids ) const

inherited

transform a vector of NodeId into a VariableeSet

Returns

NodeSet

Definition at line 160 of file graphicalModel_inl.h.

                                                              {
    VariableSet            s;
    const VariableNodeMap& v = variableNodeMap();
    for (const auto& node: l) {
      s.insert(&v.get(node));
    }
    return s;
  }

References gum::VariableNodeMap::get(), gum::Set< Key >::insert(), and variableNodeMap().

Here is the call graph for this function:

variables() [2/2]

INLINE VariableSet gum::GraphicalModel::variables ( const std::vector< std::string > & l ) const

inherited

transform a vector of names into a VariableeSet

Returns

NodeSet

Definition at line 150 of file graphicalModel_inl.h.

                                                                             {
    VariableSet            s;
    const VariableNodeMap& v = variableNodeMap();
    for (const auto& name: l) {
      s.insert(&v.variableFromName(name));
    }
    return s;
  }

References variableNodeMap().

Here is the call graph for this function:

Member Data Documentation

_propertiesMap_

HashTable< std::string, std::string > gum::GraphicalModel::_propertiesMap_

privateinherited

The properties of this Directed Graphical Model.

Definition at line 236 of file graphicalModel.h.

Referenced by _properties_(), operator=(), properties(), propertyWithDefault(), and setProperty().

graph_

UndiGraph gum::UGmodel::graph_

protectedinherited

The DAG of this Directed Graphical Model.

Definition at line 182 of file UGmodel.h.

Referenced by UGmodel(), gum::IMarkovRandomField< GUM_SCALAR >::_minimalCondSetVisit_(), edges(), exists(), existsEdge(), graph(), gum::IMarkovRandomField< GUM_SCALAR >::minimalCondSet(), neighbours(), nodes(), operator=(), and sizeEdges().

---

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

• agrum/MRF/IMarkovRandomField.h
• agrum/MRF/IMarkovRandomField_tpl.h