gum::credal::VarMod2BNsMap< GUM_SCALAR > Class Template Reference

Class used to store optimum IBayesNet during some inference algorithms. More...

#include <agrum/CN/varMod2BNsMap.h>

Collaboration diagram for gum::credal::VarMod2BNsMap< GUM_SCALAR >:

Public Member Functions
bool	insert (const std::vector< bool > &bn, const std::vector< Size > &key)
	Insert for thread fusion. More...
bool	insert (const std::vector< Size > &key, const bool isBetter)
	Thread insert. More...
Constructors / Destructors
	VarMod2BNsMap (const CredalNet< GUM_SCALAR > &cn)
	Thread constructor. More...
	VarMod2BNsMap ()
	Default constructor that should be used only by InferenceEngine since it has a member variable. More...
	~VarMod2BNsMap ()
	Destructor. More...
Getters and setters
void	setCNet (const CredalNet< GUM_SCALAR > &cn)
	Initialize __sampleDef from the CredalNet. More...
void	setCurrentSample (const std::vector< std::vector< std::vector< bool > > > &sample)
	Set the current thread sample and it's hash. More...
const dBN &	getCurrentSample ()
	Get the current sample as a vector of bits without structure. More...
const std::vector< std::vector< std::vector< bool > > > &	getSampleDef ()
	Get the sample structure. More...
const std::vector< dBN *>	getBNOptsFromKey (const std::vector< Size > &key)
	Get optimum IBayesNet (s) without structure of the given variable, modality for min or max. More...
std::vector< std::vector< std::vector< std::vector< bool > > > >	getFullBNOptsFromKey (const std::vector< Size > &key)
	Get optimum IBayesNet (s) with structure of the given variable, modality for min or max. More...
Size	getEntrySize () const
	Get the number of IBayesNet stored. More...

Protected Attributes
hashNet	_myHashNet
	Map id - dBN : get a net from it's id. More...
varHashs	_myVarHashs
	Map varKey - list(id) : get all optimal nets id from the given key : variable, modality and min=0 ( or max=1 ). More...
hashVars	_myHashVars
	Map id - list(varKey) : get all variables, modalities associated to this optimal net id. More...
std::vector< std::vector< std::vector< bool > > >	_sampleDef
	Since all samples have the same structure, this will be used as default initialization (copy constructor) for any sample. More...
dBN	_currentSample
	The current sampled IBayesNet. More...
Size	_currentHash
	The current sampled IBayesNet hash. More...
std::hash< std::vector< bool > >	_vectHash
	The hash fuction. More...
const CredalNet< GUM_SCALAR > *	cnet
	A pointer to the CredalNet to be used. More...

Detailed Description

template<typename GUM_SCALAR>
class gum::credal::VarMod2BNsMap< GUM_SCALAR >

Class used to store optimum IBayesNet during some inference algorithms.

Template Parameters

GUM_SCALAR

A floating type ( float, double, long double ... ).

Author

Matthieu HOURBRACQ and Pierre-Henri WUILLEMIN

Definition at line 56 of file varMod2BNsMap.h.

Member Typedef Documentation

dBN

template<typename GUM_SCALAR>

typedef std::vector< bool > gum::credal::VarMod2BNsMap< GUM_SCALAR >::dBN

private

Definition at line 58 of file varMod2BNsMap.h.

hashNet

template<typename GUM_SCALAR>

typedef gum::HashTable< Size, dBN > gum::credal::VarMod2BNsMap< GUM_SCALAR >::hashNet

private

Definition at line 62 of file varMod2BNsMap.h.

hashVars

template<typename GUM_SCALAR>

typedef gum::HashTable< Size, std::list< varKey > > gum::credal::VarMod2BNsMap< GUM_SCALAR >::hashVars

private

Definition at line 66 of file varMod2BNsMap.h.

varHashs

template<typename GUM_SCALAR>

typedef gum::HashTable< varKey, std::list< Size > > gum::credal::VarMod2BNsMap< GUM_SCALAR >::varHashs

private

Definition at line 64 of file varMod2BNsMap.h.

varKey

template<typename GUM_SCALAR>

typedef std::vector< Size > gum::credal::VarMod2BNsMap< GUM_SCALAR >::varKey

private

Definition at line 60 of file varMod2BNsMap.h.

Constructor & Destructor Documentation

VarMod2BNsMap() [1/2]

template<typename GUM_SCALAR >

gum::credal::VarMod2BNsMap< GUM_SCALAR >::VarMod2BNsMap ( const CredalNet< GUM_SCALAR > &  cn )

explicit

Thread constructor.

Initialize __sampleDef from the CredalNet.

Parameters

cn	The CredalNet to be used.

Definition at line 36 of file varMod2BNsMap_tpl.h.

                                                                                {
      setCNet(cn);

      GUM_CONSTRUCTOR(VarMod2BNsMap);
    }

VarMod2BNsMap() [2/2]

template<typename GUM_SCALAR >

gum::credal::VarMod2BNsMap< GUM_SCALAR >::VarMod2BNsMap

(

)

Default constructor that should be used only by InferenceEngine since it has a member variable.

Definition at line 29 of file varMod2BNsMap_tpl.h.

                                               {
      cnet = nullptr;

      GUM_CONSTRUCTOR(VarMod2BNsMap);
    }

~VarMod2BNsMap()

template<typename GUM_SCALAR >

gum::credal::VarMod2BNsMap< GUM_SCALAR >::~VarMod2BNsMap

(

)

Destructor.

Definition at line 43 of file varMod2BNsMap_tpl.h.

                                                {
      GUM_DESTRUCTOR(VarMod2BNsMap);
    }

Member Function Documentation

getBNOptsFromKey()

template<typename GUM_SCALAR >

const std::vector< std::vector< bool > *> gum::credal::VarMod2BNsMap< GUM_SCALAR >::getBNOptsFromKey

(

const std::vector< Size > &

key

)

Get optimum IBayesNet (s) without structure of the given variable, modality for min or max.

Parameters

key	The constant reference to the variable, modality, min or max.

Returns

The constant vector of not yet constant pointers to the nets.

Definition at line 198 of file varMod2BNsMap_tpl.h.

                                        {
      // return something even if key does not exist
      if (!_myVarHashs.exists(key)) return std::vector< std::vector< bool >* >();

      std::list< Size >& netsHash = _myVarHashs[key];   //.at(key);

      std::vector< dBN* > nets;
      nets.resize(netsHash.size());

      std::list< Size >::iterator it = netsHash.begin();

      for (Size i = 0; i < netsHash.size(); i++, ++it) {
        nets[i] = &_myHashNet /*.at(*/[*it];   //);
      }

      return nets;
    }

getCurrentSample()

template<typename GUM_SCALAR >

const std::vector< bool > & gum::credal::VarMod2BNsMap< GUM_SCALAR >::getCurrentSample

(

)

Get the current sample as a vector of bits without structure.

Returns

The constant reference to the sampled IBayesNet without structure.

Definition at line 186 of file varMod2BNsMap_tpl.h.

                                                                           {
      return _currentSample;
    }

getEntrySize()

template<typename GUM_SCALAR >

Size gum::credal::VarMod2BNsMap< GUM_SCALAR >::getEntrySize

(

)

const

Get the number of IBayesNet stored.

Definition at line 254 of file varMod2BNsMap_tpl.h.

                                                         {
      return _myHashNet.size();
    }

getFullBNOptsFromKey()

template<typename GUM_SCALAR >

std::vector< std::vector< std::vector< std::vector< bool > > > > gum::credal::VarMod2BNsMap< GUM_SCALAR >::getFullBNOptsFromKey

(

const std::vector< Size > &

key

)

Get optimum IBayesNet (s) with structure of the given variable, modality for min or max.

Parameters

key	The constant reference to the variable, modality, min or max.

Returns

The vector of not yet constant pointers to the nets.

Definition at line 219 of file varMod2BNsMap_tpl.h.

References GUM_ERROR.

                                        {
      if (cnet == nullptr)
        GUM_ERROR(OperationNotAllowed,
                  "No CredalNet associated to me ! Can't get FullBNOptsFromKey : "
                     << key);

      if (!_myVarHashs.exists(key))
        return std::vector< std::vector< std::vector< std::vector< bool > > > >();

      std::list< Size >& netsHash = _myVarHashs[key];   //.at(key);

      std::vector< std::vector< std::vector< std::vector< bool > > > > nets;
      nets.resize(netsHash.size(), _sampleDef);

      std::list< Size >::iterator it = netsHash.begin();

      for (Size i = 0; i < netsHash.size(); i++, ++it) {
        // std::vector< std::vector< std::vector < bool > > > net(_sampleDef);
        dBN::iterator it2 = _myHashNet /*.at(*/[*it] /*)*/.begin();

        for (Size j = 0; j < _sampleDef.size(); j++) {
          for (Size k = 0; k < _sampleDef[j].size(); k++) {
            for (Size l = 0; l < _sampleDef[j][k].size(); l++) {
              nets[i][j][k][l] = *it2;
              ++it2;
            }
          }
        }
      }

      return nets;
    }

getSampleDef()

template<typename GUM_SCALAR >

const std::vector< std::vector< std::vector< bool > > > & gum::credal::VarMod2BNsMap< GUM_SCALAR >::getSampleDef

(

)

Get the sample structure.

Returns

The constant reference to the sample structure.

Definition at line 192 of file varMod2BNsMap_tpl.h.

                                                 {
      return _sampleDef;
    }

insert() [1/2]

template<typename GUM_SCALAR >

bool gum::credal::VarMod2BNsMap< GUM_SCALAR >::insert	(	const std::vector< bool > &	bn,
		const std::vector< Size > &	key
	)

Insert for thread fusion.

All inserted nets are optimums and none will be deleted because a better one is being inserted.

Parameters

bn	The constant reference to the net to be inserted.
key	The constant reference to the key at which we will insert the net.

Definition at line 70 of file varMod2BNsMap_tpl.h.

                                                                           {
      _currentHash = Size(_vectHash(bn));
      std::list< Size >& nets =
         _myVarHashs.getWithDefault(key, std::list< Size >());   //[ key ];

      for (std::list< Size >::iterator it = nets.begin(); it != nets.end(); ++it) {
        if (*it == _currentHash) return false;
      }

      // add it
      _myHashNet.set(_currentHash, bn);   //[_currentHash] = bn;
      // insert net hash in our key net list
      nets.push_back(_currentHash);
      // insert out key in the hash key list
      _myHashVars
         .getWithDefault(_currentHash,
                         std::list< varKey >()) /*[_currentHash]*/.push_back(key);
      return true;
    }

insert() [2/2]

template<typename GUM_SCALAR >

bool gum::credal::VarMod2BNsMap< GUM_SCALAR >::insert	(	const std::vector< Size > &	key,
		const bool	isBetter
	)

Thread insert.

Parameters

key	The key at which the net will be inserted.
isBetter	True if the net is a better one, i.e. it gave better marginals.

Returns

True if the net was inserted, false otherwise.

Definition at line 92 of file varMod2BNsMap_tpl.h.

                                                                                  {
      if (isBetter) {
        // get all nets of this key (maybe entry does not exists)
        std::list< Size >& old_nets =
           _myVarHashs.getWithDefault(key, std::list< Size >());   //[ key ];

        // for each one
        for (std::list< Size >::iterator it = old_nets.begin();
             it != old_nets.end();
             ++it) {
          // get all keys associated to this net
          std::list< varKey >& netKeys =
             _myHashVars.getWithDefault(*it, std::list< varKey >());   //[ *it ];

          // if we are the sole user, delete the net entry
          if (netKeys.size() == 1) {
            _myHashVars.erase(*it);
          }
          // other keys use the net, delete our key from list
          else {
            for (std::list< varKey >::iterator it2 = netKeys.begin();
                 it2 != netKeys.end();
                 ++it2) {
              if (*it2 == key) {
                netKeys.erase(it2);
                break;
              }
            }
          }
        }   // end of : for each old_net

        // clear all old_nets
        old_nets.clear();
        // insert new net with it's hash
        _myHashNet.set(_currentHash,
                       _currentSample);   //[_currentHash] = _currentSample;
        // insert net hash in our key net list
        old_nets.push_back(_currentHash);
        // insert out key in the hash key list
        _myHashVars
           .getWithDefault(_currentHash, std::list< varKey >()) /*[_currentHash]*/
           .push_back(key);
        return true;

      }   // end of isBetter
      // another opt net
      else {
        // check that we didn't add it for this key
        std::list< Size >& nets =
           _myVarHashs.getWithDefault(key, std::list< Size >());   //[ key ];

        for (std::list< Size >::iterator it = nets.begin(); it != nets.end();
             ++it) {
          if (*it == _currentHash) return false;
        }

        // add it
        _myHashNet.set(_currentHash, _currentSample);
        // insert net hash in our key net list
        nets.push_back(_currentHash);
        // insert out key in the hash key list
        _myHashVars.getWithDefault(_currentHash, std::list< varKey >())
           .push_back(key);

        /*
              // add it
              _myHashNet[_currentHash] = _currentSample;
              // insert net hash in our key net list
              nets.push_back(_currentHash);
              // insert out key in the hash key list
              _myHashVars[_currentHash].push_back(key);
        */
        return true;
      }   // end of ! isBetter
    }

setCNet()

template<typename GUM_SCALAR >

void gum::credal::VarMod2BNsMap< GUM_SCALAR >::setCNet

(

const CredalNet< GUM_SCALAR > &

cn

)

Initialize __sampleDef from the CredalNet.

Shoud only be used by InferenceEngine to initialize it's member variable.

Parameters

cn	The CredalNet to be used.

Definition at line 48 of file varMod2BNsMap_tpl.h.

References gum::credal::CredalNet< GUM_SCALAR >::credalNet_currentCpt(), and gum::superiorPow().

                                                                               {
      auto* cpt = &cn.credalNet_currentCpt();
      auto  nNodes = cpt->size();
      _sampleDef.resize(nNodes);

      for (NodeId node = 0; node < nNodes; node++) {
        auto pConfs = (*cpt)[node].size();
        _sampleDef[node].resize(pConfs);

        for (Size pconf = 0; pconf < pConfs; pconf++) {
          Size          nVertices = Size((*cpt)[node][pconf].size());
          unsigned long b, c;   // needed by superiorPow
          superiorPow(static_cast< unsigned long >(nVertices), b, c);
          Size nBits = Size(b);
          _sampleDef[node][pconf].resize(nBits);
        }
      }

      cnet = &cn;
    }

Here is the call graph for this function:

setCurrentSample()

template<typename GUM_SCALAR >

void gum::credal::VarMod2BNsMap< GUM_SCALAR >::setCurrentSample

(

const std::vector< std::vector< std::vector< bool > > > &

sample

)

Set the current thread sample and it's hash.

Parameters

sample

The constant reference to the sample with structure.

Definition at line 170 of file varMod2BNsMap_tpl.h.

                                                                  {
      _currentSample.clear();

      for (Size i = 0; i < sample.size(); i++)
        for (Size j = 0; j < sample[j].size(); j++)
          for (Size k = 0; k < sample[i][j].size(); k++)
            _currentSample.push_back(sample[i][j][k]);

      // std::cout << sample << std::endl;
      // std::cout << _currentSample << std::endl;

      _currentHash = Size(_vectHash(_currentSample));
    }

Member Data Documentation

_currentHash

template<typename GUM_SCALAR>

Size gum::credal::VarMod2BNsMap< GUM_SCALAR >::_currentHash

protected

The current sampled IBayesNet hash.

Definition at line 87 of file varMod2BNsMap.h.

_currentSample

template<typename GUM_SCALAR>

dBN gum::credal::VarMod2BNsMap< GUM_SCALAR >::_currentSample

protected

The current sampled IBayesNet.

Definition at line 85 of file varMod2BNsMap.h.

_myHashNet

template<typename GUM_SCALAR>

hashNet gum::credal::VarMod2BNsMap< GUM_SCALAR >::_myHashNet

protected

Map id - dBN : get a net from it's id.

Definition at line 70 of file varMod2BNsMap.h.

_myHashVars

template<typename GUM_SCALAR>

hashVars gum::credal::VarMod2BNsMap< GUM_SCALAR >::_myHashVars

protected

Map id - list(varKey) : get all variables, modalities associated to this optimal net id.

Definition at line 77 of file varMod2BNsMap.h.

_myVarHashs

template<typename GUM_SCALAR>

varHashs gum::credal::VarMod2BNsMap< GUM_SCALAR >::_myVarHashs

protected

Map varKey - list(id) : get all optimal nets id from the given key : variable, modality and min=0 ( or max=1 ).

Definition at line 73 of file varMod2BNsMap.h.

_sampleDef

template<typename GUM_SCALAR>

std::vector< std::vector< std::vector< bool > > > gum::credal::VarMod2BNsMap< GUM_SCALAR >::_sampleDef

protected

Since all samples have the same structure, this will be used as default initialization (copy constructor) for any sample.

Definition at line 82 of file varMod2BNsMap.h.

_vectHash

template<typename GUM_SCALAR>

std::hash< std::vector< bool > > gum::credal::VarMod2BNsMap< GUM_SCALAR >::_vectHash

protected

The hash fuction.

Requires c++11/0x.

Definition at line 89 of file varMod2BNsMap.h.

cnet

template<typename GUM_SCALAR>

const CredalNet< GUM_SCALAR >* gum::credal::VarMod2BNsMap< GUM_SCALAR >::cnet

protected

A pointer to the CredalNet to be used.

Definition at line 92 of file varMod2BNsMap.h.

---

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

• agrum/CN/varMod2BNsMap.h
• agrum/CN/varMod2BNsMap_tpl.h