gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber > Class Template Reference

<agrum/BN/generator/MCayesNetGenerator.h> More...

#include <MCBayesNetGenerator.h>

Inheritance diagram for gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >:

Collaboration diagram for gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >:

Public Member Functions
Constructors / Destructor
	MCBayesNetGenerator (Size nbrNodes, Size maxArcs, Idx maxModality=2, Size iteration=5000, Idx p=30, Idx q=40)
	Constructor. More...
	MCBayesNetGenerator (BayesNet< GUM_SCALAR > bayesNet, Size iteration=5000, Idx p=30, Idx q=40)
	Constructor. More...
	~MCBayesNetGenerator () override
	Destructor. More...
BN generation methods
void	generateBN (BayesNet< GUM_SCALAR > &bayesNet) override
	Generates a random Bayesian network. More...
void	disturbBN (BayesNet< GUM_SCALAR > &bayesNetinit, Size iteration=0)
	Change randomly the topology of a specific Bayesian networks. More...
Getters
Size	iteration () const
	Return a constant reference to the number of iteration imposed on the Markov Chain BayesNetGenerator. More...
Idx	p () const
	Return a constant reference to the probabilité p imposed on the Markov Chain BayesNetGenerator. More...
Idx	q () const
	Return a constant reference to the probabilité imposed on the Markov Chain BayesNetGenerator. More...
Setters
void	setIteration (Size iteration)
	Modifies the value of the number of iterations impose on the BayesNetGenerator. More...
void	setP (Idx p)
	Modifies the value of the probability p imposed on the BayesNetGenerator. More...
void	setQ (Idx q)
	Modifies the value of the probability q imposed on the BayesNetGenerator. More...
BN generation methods
void	fillCPT ()
	function that insert random values in the CPT of each nodes according to the CPTGenerator. More...
Getters
Size	nbrNodes () const
	Return a constant reference to the number of nodes imposed on the IBayesNetGenerator. More...
Size	maxArcs () const
	Return a constant reference to the maximum number of arcs imposed on the IBayesNetGenerator. More...
Size	maxModality () const
	Return a constant reference to the maximum modality imposed on the IBayesNetGenerator. More...
Setters
void	setNbrNodes (Size nbrNodes)
	Modifies the value of the number of nodes imposed on the BayesGenerator. More...
void	setMaxArcs (Size maxArcs)
	Modifies the value of the number of nodes imposed on the BayesGenerator. More...
void	setMaxModality (Size maxModality)
	Modifies the value of the number of nodes imposed on the BayesGenerator. More...

Protected Attributes
Size	iteration_
Idx	p_
Idx	q_
bool	disturbing_
BayesNet< GUM_SCALAR >	bayesNettemp_
HashTable< NodeId, Potential< GUM_SCALAR > *>	hashMarginal_
Size	nbrNodes_
Size	maxArcs_
Size	maxModality_
BayesNet< GUM_SCALAR >	bayesNet_

Protected Member Functions
bool	isPolytree__ ()
	The function that verify if graph is a polytree. More...
bool	connect__ (NodeId i, NodeId j)
	The function that verify if node i and j are connected. More...
bool	directedPath__ (NodeId tail, NodeId head)
	The function that verify if there is a oriented path from node i to node j. More...
void	insertArc__ (NodeId i, NodeId j)
	The function that will insert an arc between node i to node j, but only if there isn't any cycle created. More...
void	eraseArc__ (NodeId i, NodeId j, bool mustbeconnex=true)
	The function that will remove the arc between node i and node j. More...
void	PMMx_poly__ ()
	In the case that the graph is a polytree, the function will, according to the probability p and q, choose which change of state must occure( AorR or AR or jump) then will assert that the imposed constraint are respected and if not, will return to the previous topology. More...
void	PMMx_multi__ ()
	In the case that the graph is a multiconnected graph, the function will, according to the probability p and q, choose which change of state must occure( AorR or jump) then will assert that the imposed constraint are respected and if not, will return to the previous topology. More...
void	jump_poly__ ()
	In the case that the graph is a polytree, the function will add a ramdom arc by the use of the function insertArc__ if the arc does not exist allready. More...
void	jump_multi__ ()
	In the case that the graph is a multiconnect graph, the function will choose randomly two nodes and will remove the arc between them by the use of the function insertArc__ if the arc exists. More...
void	AorR__ ()
	The function will add or remove a random arc in the graph using the functions insertArc__ and removeArc__. More...
void	AR__ ()
	The function will remove and add a random arc changing the topology of the graph but asserting its connectivity. More...
virtual bool	checkConditions__ ()
	The boolean function that will assert the respect of the constraint. More...
bool	connect__ (NodeId i, NodeId j, NodeSet &excluded)
	The internal function used by the previous connect__. More...
bool	directedPath__ (NodeId tail, NodeId head, NodeSet &excluded)
	The internal function used by the previous directedPath__. More...
void	chooseNodes__ (NodeId &i, NodeId &j)
	The function that randomly choose two nodes of the graph. More...
void	chooseCloseNodes__ (NodeId &i, NodeId &j)
	The function that randomly choose two neighbours nodes of the graph. More...
void	transformPoly__ (Idx nbiter)
	The function that randomly change the simple tree into a polytree. More...
void	createTree__ (Size BNSize)
	The function that randomly generate a simple tree. More...
NodeId	createPartTree__ (Size BNSize, Idx &n)
	The internal function used by createTree__ that randomly generate a simple tree. More...

Detailed Description

template<typename GUM_SCALAR, template< typename > class ICPTGenerator = SimpleCPTGenerator, template< typename > class ICPTDisturber = SimpleCPTDisturber>
class gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >

<agrum/BN/generator/MCayesNetGenerator.h>

Class for generating Bayesian networks with Markov chains.

This class randomly generates a Bayesian network given 6 parameters: the number of nodes, the maximum number of arcs the and of iterations the maximum modality.

Warning

Be Careful when entering the parameters, high Values may cause the density of the Bayesian network to be too high resulting in the failure of most of the inference Methods.

This Generation method require the use of two probability parameters(p, q) defining the choice of processes that will provoke the change of state. You can see in the graph below how the probabilities are used.

Definition at line 159 of file MCBayesNetGenerator.h.

Constructor & Destructor Documentation

MCBayesNetGenerator() [1/2]

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::MCBayesNetGenerator	(	Size	nbrNodes,
		Size	maxArcs,
		Idx	maxModality = 2,
		Size	iteration = 5000,
		Idx	p = 30,
		Idx	q = 40
	)

Constructor.

Use by default the SimpleCPTGenerator for generating the BNs CPT and the SimpleCPTDisturber to tweak the CPT when the dimension of the table changes.

Parameters

nbrNodes	The number of nodes in the generated BN.
maxArcs	The maximum number of Arcs.
maxModality	Each DRV has from 2 to maxModality modalities
iteration	The number of iterations wanted to repeat the algorithm
p	probability for the change of the state (see use of p and q )
q	probability for the change of the state (see use of p and q )

Definition at line 58 of file MCBayesNetGenerator_tpl.h.

                                 :
      IBNG(nbrNodes, maxArcs, maxModality),
      bayesNettemp_(), hashMarginal_() {
    if (p + q > 100)
      GUM_ERROR(
         OperationNotAllowed,
         "the sum of the probabilities p and q must be at most equal to 100");

    iteration_  = iteration;
    p_          = p;
    q_          = q;
    disturbing_ = false;

    GUM_CONSTRUCTOR(MCBayesNetGenerator);
  }

MCBayesNetGenerator() [2/2]

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::MCBayesNetGenerator ( BayesNet< GUM_SCALAR >  bayesNet, Size  iteration = 5000, Idx  p = 30, Idx  q = 40  )

explicit

Constructor.

Use by default the SimpleCPTGenerator for generating the BNs CPT and the SimpleCPTDisturber to tweak the CPT when the dimension of the table changes.

Parameters

bayesNet	the IBayesNet used as reference to fill the parameters nbrNodes, maxArcs and maxModality
iteration	The number of iterations wanted to repeat the algorithm
p	probability for the change of the state (see probability_p_q )
q	probability for the change of the state (see probability_p_q )

Definition at line 85 of file MCBayesNetGenerator_tpl.h.

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

                                                   :
      MCBayesNetGenerator(bayesNet.size(),
                          (Size)(bayesNet.sizeArcs() * 1.1),
                          getMaxModality(bayesNet)) {
    iteration_  = iteration;
    p_          = p;
    q_          = q;
    disturbing_ = false;
  }

Here is the call graph for this function:

~MCBayesNetGenerator()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::~MCBayesNetGenerator ( )

override

Destructor.

Definition at line 105 of file MCBayesNetGenerator_tpl.h.

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

                            {
    GUM_DESTRUCTOR(MCBayesNetGenerator);
  }

Here is the call graph for this function:

Member Function Documentation

AorR__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::AorR__ ( )

protected

The function will add or remove a random arc in the graph using the functions insertArc__ and removeArc__.

Definition at line 294 of file MCBayesNetGenerator_tpl.h.

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

                                                                               {
    NodeId i, j;
    chooseNodes__(i, j);
    const DAG dag__ = this->bayesNet_.dag();

    if (dag__.existsArc(i, j)) {
      eraseArc__(i, j);

      return;
    } else
      insertArc__(i, j);
  }

Here is the call graph for this function:

AR__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::AR__ ( )

protected

The function will remove and add a random arc changing the topology of the graph but asserting its connectivity.

Definition at line 312 of file MCBayesNetGenerator_tpl.h.

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

                                                                             {
    NodeId i, j, head, tail;
    chooseNodes__(i, j);
    const DAG dag__ = this->bayesNet_.dag();

    if (dag__.existsArc(i, j) || dag__.existsArc(j, i)) {
      return;
    } else {
      Idx per = randomValue(100);

      if (per < 50) {
        head = i;
        tail = j;
      } else {
        head = j;
        tail = i;
      }

      for (auto node: dag__.parents(j)) {
        NodeSet excluded;
        excluded.insert(j);

        if (connect__(node, i, excluded)) {
          this->bayesNet_.eraseArc(node, j);
          this->bayesNet_.addArc(head, tail);
          return;
        }
      }

      for (auto node: dag__.children(j)) {
        NodeSet excluded;
        excluded.insert(j);

        if (connect__(node, i, excluded)) {
          this->bayesNet_.eraseArc(j, node);
          this->bayesNet_.addArc(head, tail);
          return;
        }
      }
    }
  }

Here is the call graph for this function:

checkConditions__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE bool gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::checkConditions__ ( )

protectedvirtual

The boolean function that will assert the respect of the constraint.

Reimplemented in gum::MaxParentsMCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >, and gum::MaxInducedWidthMCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >.

Definition at line 187 of file MCBayesNetGenerator_tpl.h.

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

                         {
    return this->maxArcs_ >= this->bayesNet_.sizeArcs();
  }

Here is the call graph for this function:

chooseCloseNodes__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::chooseCloseNodes__ ( NodeId &  i, NodeId &  j  )

protected

The function that randomly choose two neighbours nodes of the graph.

Definition at line 451 of file MCBayesNetGenerator_tpl.h.

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

                                              {
    NodeId temp = randomValue(this->bayesNet_.size());
    Size   co   = 0;

    if (this->bayesNet_.parents(temp).size()) {
      j       = temp;
      auto it = this->bayesNet_.parents(j).begin();
      co      = randomValue(this->bayesNet_.parents(j).size());

      while (co--) {
        ++it;
      }

      i = *it;
    } else if (this->bayesNet_.children(temp).size()) {
      i       = temp;
      auto it = this->bayesNet_.children(i).begin();
      co      = randomValue(this->bayesNet_.children(i).size());

      while (co--) {
        ++it;
      }

      j = *it;
    } else {
      GUM_ERROR(FatalError, "Sorry Misconstructed BN because of isolated node.");
    }
  }

Here is the call graph for this function:

chooseNodes__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::chooseNodes__ ( NodeId &  i, NodeId &  j  )

protected

The function that randomly choose two nodes of the graph.

Definition at line 437 of file MCBayesNetGenerator_tpl.h.

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

                                         {
    i = randomValue(this->bayesNet_.size());
    j = randomValue(this->bayesNet_.size());

    while (i == j)
      j = randomValue(this->bayesNet_.size());
  }

Here is the call graph for this function:

connect__() [1/2]

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

bool gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::connect__ ( NodeId  i, NodeId  j  )

protected

The function that verify if node i and j are connected.

Definition at line 565 of file MCBayesNetGenerator_tpl.h.

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

                     {
    const DAG dag__ = this->bayesNet_.dag();

    if (dag__.existsArc(i, j) || dag__.existsArc(j, i))
      return true;
    else {
      NodeSet excluded;
      excluded.insert(i);

      for (auto par: dag__.parents(i)) {
        if (!excluded.exists(par) && connect__(par, j, excluded)) return true;
      }

      for (auto chi: dag__.children(i)) {
        if (!excluded.exists(chi) && connect__(chi, j, excluded)) return true;
      }

      return false;
    }
  }

Here is the call graph for this function:

connect__() [2/2]

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

bool gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::connect__ ( NodeId  i, NodeId  j, NodeSet &  excluded  )

protected

The internal function used by the previous connect__.

It asserts the existence of an unoriented path between node i and node j avoiding passing through nodes listed in excluded.

Definition at line 593 of file MCBayesNetGenerator_tpl.h.

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

                            {
    const DAG dag__ = this->bayesNet_.dag();

    if (dag__.existsArc(i, j) || dag__.existsArc(j, i))
      return true;
    else {
      excluded.insert(i);

      for (auto par: dag__.parents(i)) {
        if (!excluded.exists(par) && connect__(par, j, excluded)) return true;
      }

      for (auto chi: dag__.children(i)) {
        if (!excluded.exists(chi) && connect__(chi, j, excluded)) return true;
      }

      return false;
    }
  }

Here is the call graph for this function:

createPartTree__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

NodeId gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::createPartTree__ ( Size  BNSize, Idx &  n  )

protected

The internal function used by createTree__ that randomly generate a simple tree.

n : id number for node label

Definition at line 511 of file MCBayesNetGenerator_tpl.h.

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

                                           {
    Size              nb_mod = 2 + randomValue(this->maxModality_ - 1);
    std::stringstream strBuff;
    strBuff << "n_" << n++;
    NodeId root
       = this->bayesNet_.add(LabelizedVariable(strBuff.str(), "", nb_mod));
    Size maxNodes = BNSize - 1;
    Size SubG     = 0;

    while (maxNodes) {
      SubG         = randomValue(maxNodes) + 1;
      maxNodes     = maxNodes - SubG;
      NodeId rootS = createPartTree__(SubG, n);
      this->bayesNet_.addArc(root, rootS);
    }

    return root;
  }

Here is the call graph for this function:

createTree__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::createTree__ ( Size  BNSize )

protected

The function that randomly generate a simple tree.

Definition at line 486 of file MCBayesNetGenerator_tpl.h.

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

                     {
    Idx               n      = 0;
    Size              nb_mod = 2 + randomValue(this->maxModality_ - 1);
    std::stringstream strBuff;
    strBuff << "n_" << n++;
    NodeId root
       = this->bayesNet_.add(LabelizedVariable(strBuff.str(), "", nb_mod));
    Size maxNodes = BNSize - 1;
    Size SubG     = 0;

    while (maxNodes) {
      SubG         = randomValue(maxNodes) + 1;
      maxNodes     = maxNodes - SubG;
      NodeId rootS = createPartTree__(SubG, n);
      this->bayesNet_.addArc(root, rootS);
    }
  }

Here is the call graph for this function:

directedPath__() [1/2]

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

bool gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::directedPath__ ( NodeId  tail, NodeId  head  )

protected

The function that verify if there is a oriented path from node i to node j.

Definition at line 622 of file MCBayesNetGenerator_tpl.h.

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

                                              {
    const DAG dag__ = this->bayesNet_.dag();

    if (dag__.existsArc(tail, head))
      return true;
    else {
      NodeSet excluded;
      excluded.insert(tail);

      for (auto node: dag__.children(tail)) {
        if (directedPath__(node, head, excluded)) return true;
      }

      return false;
    }
  }

Here is the call graph for this function:

directedPath__() [2/2]

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

bool gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::directedPath__ ( NodeId  tail, NodeId  head, NodeSet &  excluded  )

protected

The internal function used by the previous directedPath__.

It asserts the existence of an oriented path between node i and node j avoiding passing through nodes listed in excluded.

Definition at line 645 of file MCBayesNetGenerator_tpl.h.

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

                                                                 {
    const DAG dag__ = this->bayesNet_.dag();

    if (dag__.existsArc(tail, head))
      return true;
    else {
      excluded.insert(tail);

      for (auto node: dag__.children(tail)) {
        if (!excluded.exists(node) && directedPath__(node, head, excluded))
          return true;
      }

      return false;
    }
  }

Here is the call graph for this function:

disturbBN()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::disturbBN	(	BayesNet< GUM_SCALAR > &	bayesNetinit,
		Size	iteration = 0
	)

Change randomly the topology of a specific Bayesian networks.

Parameters

bayesNetinit	IBayesNet to be modify
iteration	The number of iterations wanted to repeat the algorithm

Returns

null but modify inputed Bayesian network

Exceptions

OperationNotAllow

if the initial state of the IBayesNet is not respecting the wanted conditions if iteration = 0, it is assumed that the number of iteration wanted is the same as the one specified in the constructor

Definition at line 136 of file MCBayesNetGenerator_tpl.h.

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

                                        {   // insert option for the variation
    disturbing_ = true;
    Size iter   = iteration_;

    if (iteration) iteration_ = iteration;

    this->bayesNet_ = bayesNetinit;

    if (checkConditions__()) {
      LazyPropagation< GUM_SCALAR > inf(&bayesNetinit);
      inf.makeInference();

      for (auto node: bayesNetinit.nodes()) {
        auto pottemp = new Potential< GUM_SCALAR >();
        pottemp->copy(inf.posterior(node));
        hashMarginal_.insert(node, pottemp);
      }

      bayesNettemp_ = this->bayesNet_;

      if (isPolytree__())
        PMMx_poly__();
      else
        PMMx_multi__();

      bayesNetinit = (this->bayesNet_);

      while (hashMarginal_.size()) {
        delete (hashMarginal_.begin().val());
        hashMarginal_.erase(
           hashMarginal_.beginSafe());   // safe iterator needed here.
      }

    } else {
      std::cout << this->bayesNet_.toDot() << std::endl;
      GUM_ERROR(OperationNotAllowed,
                "BN is not valid cause it does not respect constraint ");
    }

    iteration_  = iter;
    disturbing_ = false;
  }

Here is the call graph for this function:

eraseArc__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::eraseArc__ ( NodeId  i, NodeId  j, bool  mustbeconnex = true  )

protected

The function that will remove the arc between node i and node j.

If the boolean parameter mustbeconnex is true, the function will assert that the graph remain connected and will restore the arc otherwise.

Definition at line 408 of file MCBayesNetGenerator_tpl.h.

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

                             {
    if (disturbing_) {
      const BayesNet< GUM_SCALAR > bayesNet(this->bayesNet_);
      Potential< GUM_SCALAR >      potj;
      potj.copy(this->bayesNet_.cpt(j));
      this->bayesNet_.eraseArc(i, j);

      if (connect__(i, j) || !mustbeconnex) {
        auto marg = *hashMarginal_[i];

        this->disturbReducCPT(j, this->bayesNet_, potj, marg);
      } else
        this->bayesNet_.addArc(i, j);
    } else {
      this->bayesNet_.eraseArc(i, j);

      if (!connect__(i, j) && mustbeconnex) { this->bayesNet_.addArc(i, j); }
    }
  }

Here is the call graph for this function:

fillCPT()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

void gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::fillCPT ( )

inherited

function that insert random values in the CPT of each nodes according to the CPTGenerator.

Returns

null but modify inputed empty Bayesian network

Definition at line 64 of file IBayesNetGenerator_tpl.h.

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

                                                                {
    for (auto node: bayesNet_.nodes())
      this->generateCPT(bayesNet_.cpt(node).pos(bayesNet_.variable(node)),
                        bayesNet_.cpt(node));
  }

Here is the call graph for this function:

generateBN()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::generateBN ( BayesNet< GUM_SCALAR > &  bayesNet )

overridevirtual

Generates a random Bayesian network.

Parameters

bayesNet

empty IBayesNet to generate.

Returns

null but modify inputed Bayesian network

Implements gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >.

Definition at line 114 of file MCBayesNetGenerator_tpl.h.

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

                                       {
    Idx iteration = iteration_;

    // this->bayesNet_ = bayesNet;
    createTree__(this->nbrNodes_);
    transformPoly__(this->nbrNodes_ / 2);
    bayesNettemp_ = this->bayesNet_;
    PMMx_poly__();

    this->fillCPT();
    iteration_ = iteration;

    bayesNet = this->bayesNet_;
  }

Here is the call graph for this function:

insertArc__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::insertArc__ ( NodeId  i, NodeId  j  )

protected

The function that will insert an arc between node i to node j, but only if there isn't any cycle created.

Definition at line 388 of file MCBayesNetGenerator_tpl.h.

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

                  {
    if (directedPath__(j, i)) return;

    if (disturbing_) {
      auto potj = this->bayesNet_.cpt(j);
      this->bayesNet_.addArc(i, j);

      this->disturbAugmCPT(j, this->bayesNet_, potj, (GUM_SCALAR)0.5);
    } else
      this->bayesNet_.addArc(i, j);
  }

Here is the call graph for this function:

isPolytree__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE bool gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::isPolytree__ ( )

protected

The function that verify if graph is a polytree.

Definition at line 555 of file MCBayesNetGenerator_tpl.h.

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

                    {
    const DAG dag__ = this->bayesNet_.dag();
    return this->bayesNet_.size() - 1 == this->bayesNet_.sizeArcs();
  }

Here is the call graph for this function:

iteration()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE Size gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::iteration

(

)

const

Return a constant reference to the number of iteration imposed on the Markov Chain BayesNetGenerator.

Definition at line 668 of file MCBayesNetGenerator_tpl.h.

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

              {
    return iteration_;
  }

Here is the call graph for this function:

jump_multi__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::jump_multi__ ( )

protected

In the case that the graph is a multiconnect graph, the function will choose randomly two nodes and will remove the arc between them by the use of the function insertArc__ if the arc exists.

Definition at line 374 of file MCBayesNetGenerator_tpl.h.

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

                    {
    NodeId i, j;
    chooseNodes__(i, j);
    const DAG dag__ = this->bayesNet_.dag();

    if (dag__.existsArc(i, j)) { eraseArc__(i, j); }
  }

Here is the call graph for this function:

jump_poly__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::jump_poly__ ( )

protected

In the case that the graph is a polytree, the function will add a ramdom arc by the use of the function insertArc__ if the arc does not exist allready.

Definition at line 360 of file MCBayesNetGenerator_tpl.h.

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

                   {
    NodeId i, j;
    chooseNodes__(i, j);
    const DAG dag__ = this->bayesNet_.dag();

    if (!dag__.existsArc(i, j)) insertArc__(i, j);
  }

Here is the call graph for this function:

maxArcs()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

INLINE Size gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::maxArcs ( ) const

inherited

Return a constant reference to the maximum number of arcs imposed on the IBayesNetGenerator.

Definition at line 82 of file IBayesNetGenerator_tpl.h.

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

                                                                             {
    return maxArcs_;
  }

Here is the call graph for this function:

maxModality()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

INLINE Size gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::maxModality ( ) const

inherited

Return a constant reference to the maximum modality imposed on the IBayesNetGenerator.

Definition at line 72 of file IBayesNetGenerator_tpl.h.

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

                                                                        {
    return maxModality_;
  }

Here is the call graph for this function:

nbrNodes()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

INLINE Size gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::nbrNodes ( ) const

inherited

Return a constant reference to the number of nodes imposed on the IBayesNetGenerator.

Definition at line 77 of file IBayesNetGenerator_tpl.h.

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

                                                                              {
    return nbrNodes_;
  }

Here is the call graph for this function:

p()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE Idx gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::p

(

)

const

Return a constant reference to the probabilité p imposed on the Markov Chain BayesNetGenerator.

Definition at line 679 of file MCBayesNetGenerator_tpl.h.

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

                                                                              {
    return p_;
  }

Here is the call graph for this function:

PMMx_multi__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::PMMx_multi__ ( )

protected

In the case that the graph is a multiconnected graph, the function will, according to the probability p and q, choose which change of state must occure( AorR or jump) then will assert that the imposed constraint are respected and if not, will return to the previous topology.

Definition at line 246 of file MCBayesNetGenerator_tpl.h.

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

                    {
    if (!iteration_--) return;

    Idx per = randomValue(100);

    if (per < p_ + q_) {
      AorR__();

      if (checkConditions__()) {
        if (isPolytree__()) {
          if (per < p_) {
            bayesNettemp_ = this->bayesNet_;
            PMMx_poly__();
          } else {
            this->bayesNet_ = bayesNettemp_;
            PMMx_multi__();
          }
        } else {
          bayesNettemp_ = this->bayesNet_;
          PMMx_multi__();
        }
      } else {
        this->bayesNet_ = bayesNettemp_;
        PMMx_multi__();
      }
    } else {
      jump_multi__();

      if (checkConditions__()) {
        bayesNettemp_ = this->bayesNet_;

        if (isPolytree__())
          PMMx_poly__();
        else
          PMMx_multi__();

      } else {
        this->bayesNet_ = bayesNettemp_;
        PMMx_multi__();
      }
    }
  }

Here is the call graph for this function:

PMMx_poly__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::PMMx_poly__ ( )

protected

In the case that the graph is a polytree, the function will, according to the probability p and q, choose which change of state must occure( AorR or AR or jump) then will assert that the imposed constraint are respected and if not, will return to the previous topology.

Definition at line 200 of file MCBayesNetGenerator_tpl.h.

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

                   {
    if (!iteration_--) return;

    Idx per = randomValue(100);

    if (per < p_) {
      AorR__();

      if (checkConditions__()) {
        bayesNettemp_ = this->bayesNet_;
        PMMx_multi__();
      } else {
        this->bayesNet_ = bayesNettemp_;
        PMMx_poly__();
      }
    } else {
      if (per < p_ + q_) {
        AR__();

        if (!checkConditions__()) {
          this->bayesNet_ = bayesNettemp_;
        } else
          bayesNettemp_ = this->bayesNet_;

        PMMx_poly__();
      } else {
        jump_poly__();

        if (checkConditions__()) {
          bayesNettemp_ = this->bayesNet_;
          PMMx_multi__();

        } else {
          this->bayesNet_ = bayesNettemp_;
          PMMx_poly__();
        }
      }
    }
  }

Here is the call graph for this function:

q()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE Idx gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::q

(

)

const

Return a constant reference to the probabilité imposed on the Markov Chain BayesNetGenerator.

Definition at line 689 of file MCBayesNetGenerator_tpl.h.

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

                                                                              {
    return q_;
  }

Here is the call graph for this function:

setIteration()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::setIteration

(

Size

iteration

)

Modifies the value of the number of iterations impose on the BayesNetGenerator.

Definition at line 699 of file MCBayesNetGenerator_tpl.h.

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

                        {
    iteration_ = iteration;
  }

Here is the call graph for this function:

setMaxArcs()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

INLINE void gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::setMaxArcs ( Size  maxArcs )

inherited

Modifies the value of the number of nodes imposed on the BayesGenerator.

Definition at line 106 of file IBayesNetGenerator_tpl.h.

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

                                                                             {
    if (maxArcs < nbrNodes_ - 1 || maxArcs > (nbrNodes_ * (nbrNodes_ - 1)) / 2)
      GUM_ERROR(OperationNotAllowed, " maxArcs value not possible ");

    maxArcs_ = maxArcs;
  }

Here is the call graph for this function:

setMaxModality()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

INLINE void gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::setMaxModality ( Size  maxModality )

inherited

Modifies the value of the number of nodes imposed on the BayesGenerator.

Definition at line 87 of file IBayesNetGenerator_tpl.h.

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

                       {
    if (maxModality < 2)
      GUM_ERROR(OperationNotAllowed,
                " maxModality must be at least equal to two ");

    maxModality_ = maxModality;
  }

Here is the call graph for this function:

setNbrNodes()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

INLINE void gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::setNbrNodes ( Size  nbrNodes )

inherited

Modifies the value of the number of nodes imposed on the BayesGenerator.

Definition at line 97 of file IBayesNetGenerator_tpl.h.

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

                                                                               {
    if ((maxArcs_ < nbrNodes - 1) || (maxArcs_ > (nbrNodes * (nbrNodes - 1)) / 2))
      GUM_ERROR(OperationNotAllowed, " nbrNodes value not possible ");

    nbrNodes_ = nbrNodes;
  }

Here is the call graph for this function:

setP()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::setP

(

Idx

p

)

Modifies the value of the probability p imposed on the BayesNetGenerator.

Definition at line 709 of file MCBayesNetGenerator_tpl.h.

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

                                                                                {
    p_ = p;

    if (p + q_ > 100)
      GUM_ERROR(
         OperationNotAllowed,
         "the sum of the probabilities p and q must be at most equal to 100");
  }

Here is the call graph for this function:

setQ()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

INLINE void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::setQ

(

Idx

q

)

Modifies the value of the probability q imposed on the BayesNetGenerator.

Definition at line 723 of file MCBayesNetGenerator_tpl.h.

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

                                                                                {
    q_ = q;

    if (p_ + q > 100)
      GUM_ERROR(
         OperationNotAllowed,
         "the sum of the probabilities p and q must be at most equal to 100");
  }

Here is the call graph for this function:

transformPoly__()

template<typename GUM_SCALAR , template< typename > class ICPTGenerator, template< typename > class ICPTDisturber>

void gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::transformPoly__ ( Idx  nbiter )

protected

The function that randomly change the simple tree into a polytree.

Definition at line 537 of file MCBayesNetGenerator_tpl.h.

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

                                 {
    while (nbiter--) {
      NodeId i, j;
      chooseCloseNodes__(i, j);
      bayesNettemp_ = this->bayesNet_;
      eraseArc__(i, j, false);
      this->bayesNet_.addArc(j, i);

      if (!checkConditions__()) this->bayesNet_ = bayesNettemp_;
    }
  }

Here is the call graph for this function:

Member Data Documentation

bayesNet_

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

BayesNet< GUM_SCALAR > gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::bayesNet_

protectedinherited

Definition at line 166 of file IBayesNetGenerator.h.

bayesNettemp_

template<typename GUM_SCALAR , template< typename > class ICPTGenerator = SimpleCPTGenerator, template< typename > class ICPTDisturber = SimpleCPTDisturber>

BayesNet< GUM_SCALAR > gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::bayesNettemp_

protected

Definition at line 289 of file MCBayesNetGenerator.h.

disturbing_

template<typename GUM_SCALAR , template< typename > class ICPTGenerator = SimpleCPTGenerator, template< typename > class ICPTDisturber = SimpleCPTDisturber>

bool gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::disturbing_

protected

Definition at line 288 of file MCBayesNetGenerator.h.

hashMarginal_

template<typename GUM_SCALAR , template< typename > class ICPTGenerator = SimpleCPTGenerator, template< typename > class ICPTDisturber = SimpleCPTDisturber>

HashTable< NodeId, Potential< GUM_SCALAR >* > gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::hashMarginal_

protected

Definition at line 290 of file MCBayesNetGenerator.h.

iteration_

template<typename GUM_SCALAR , template< typename > class ICPTGenerator = SimpleCPTGenerator, template< typename > class ICPTDisturber = SimpleCPTDisturber>

Size gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::iteration_

protected

Definition at line 286 of file MCBayesNetGenerator.h.

maxArcs_

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

Size gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::maxArcs_

protectedinherited

Definition at line 164 of file IBayesNetGenerator.h.

maxModality_

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

Size gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::maxModality_

protectedinherited

Definition at line 165 of file IBayesNetGenerator.h.

nbrNodes_

template<typename GUM_SCALAR , template< typename > class ICPTGenerator>

Size gum::IBayesNetGenerator< GUM_SCALAR, ICPTGenerator >::nbrNodes_

protectedinherited

Definition at line 163 of file IBayesNetGenerator.h.

p_

template<typename GUM_SCALAR , template< typename > class ICPTGenerator = SimpleCPTGenerator, template< typename > class ICPTDisturber = SimpleCPTDisturber>

Idx gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::p_

protected

Definition at line 287 of file MCBayesNetGenerator.h.

q_

template<typename GUM_SCALAR , template< typename > class ICPTGenerator = SimpleCPTGenerator, template< typename > class ICPTDisturber = SimpleCPTDisturber>

Idx gum::MCBayesNetGenerator< GUM_SCALAR, ICPTGenerator, ICPTDisturber >::q_

protected

Definition at line 287 of file MCBayesNetGenerator.h.

---

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

• agrum/BN/generator/MCBayesNetGenerator.h
• agrum/BN/generator/MCBayesNetGenerator_tpl.h