void ExactInf::construct() {
// clear variable beliefs and reserve space
_beliefsV.clear();
_beliefsV.reserve( nrVars() );
for( size_t i = 0; i < nrVars(); i++ )
_beliefsV.push_back( Factor( var(i) ) );
// clear factor beliefs and reserve space
_beliefsF.clear();
_beliefsF.reserve( nrFactors() );
for( size_t I = 0; I < nrFactors(); I++ )
_beliefsF.push_back( Factor( factor(I).vars() ) );
}
Factor ExactInf::belief( const VarSet &ns ) const {
if( ns.size() == 0 )
return Factor();
else if( ns.size() == 1 ) {
return beliefV( findVar( *(ns.begin()) ) );
} else {
size_t I;
for( I = 0; I < nrFactors(); I++ )
if( factor(I).vars() >> ns )
break;
if( I == nrFactors() )
DAI_THROW(BELIEF_NOT_AVAILABLE);
return beliefF(I).marginal(ns);
}
}
void CBP::construct() {
// prepare datastructures for compression
for (size_t i=0; i<nrVars(); i++) {
_gndVarToSuperVar[i] = i;
}
for (size_t i=0; i<nrFactors(); i++) {
_gndFacToSuperFac[i] = i;
}
// create edge properties
_edges.clear();
_edges.reserve( nrVars() );
for( size_t i = 0; i < nrVars(); ++i ) {
_edges.push_back( vector<EdgeProp>() );
_edges[i].reserve( nbV(i).size() );
foreach( const Neighbor &I, nbV(i) ) {
EdgeProp newEP;
size_t edgeCount = factor(I).counts()[I.dual].size();
newEP.message = vector<Prob>(edgeCount,Prob( var(i).states() ));
newEP.newMessage = vector<Prob>(edgeCount,Prob( var(i).states() ));
newEP.count = vector<int>(edgeCount);
newEP.index = vector<ind_t>(edgeCount);
newEP.nrPos = edgeCount;
// simulate orginal varSet with possibly more variables, must ensure that the number of variables is equal to number of ground variables
VarSet gndVarSet;
size_t varCount = 0;
foreach(const Neighbor &tmpVar, nbF(I)) {
for(map<size_t, int>::const_iterator iter=factor(I).counts()[tmpVar.iter].begin(); iter!=factor(I).counts()[tmpVar.iter].end();iter++) {
gndVarSet |= Var(varCount, var(tmpVar).states());
varCount++;
}
}
varCount = 0;
foreach(const Neighbor &tmpVar, nbF(I)) {
size_t pos=0;
for(map<size_t, int>::const_iterator iter=factor(I).counts()[tmpVar.iter].begin(); iter!=factor(I).counts()[tmpVar.iter].end();iter++,pos++) {
if (tmpVar == i) {
// assumes that counts are iterated in increases order of positions
size_t sortedPos = factor(I).sigma()[(*iter).first];
newEP.count[pos] = (*iter).second;
newEP.index[pos].reserve( factor(I).states() );
for( IndexFor k( Var(sortedPos, var(i).states()), gndVarSet ); k >= 0; ++k ) {
newEP.index[pos].push_back( k );
}
}
varCount++;
}
}
_edges[i].push_back( newEP );
}
Real ExactInf::run() {
if( props.verbose >= 1 )
cerr << "Starting " << identify() << "...";
Factor P;
for( size_t I = 0; I < nrFactors(); I++ )
P *= factor(I);
Real Z = P.sum();
_logZ = std::log(Z);
for( size_t i = 0; i < nrVars(); i++ )
_beliefsV[i] = P.marginal(var(i));
for( size_t I = 0; I < nrFactors(); I++ )
_beliefsF[I] = P.marginal(factor(I).vars());
if( props.verbose >= 1 )
cerr << "finished" << endl;
return 0.0;
}
void CFactorGraph::printDot( std::ostream &os ) const {
os << "graph G {" << endl;
os << "node[shape=circle,width=0.4,fixedsize=true];" << endl;
for( size_t i = 0; i < nrVars(); i++ )
os << "\tv" << var(i).label() << ";" << endl;
os << "node[shape=box,width=0.3,height=0.3,fixedsize=true];" << endl;
for( size_t I = 0; I < nrFactors(); I++ )
os << "\tf" << I << ";" << endl;
for( size_t i = 0; i < nrVars(); i++ )
foreach( const Neighbor &I, nbV(i) ) // for all neighboring factors I of i
os << "\tv" << var(i).label() << " -- f" << I << ";" << endl;
os << "}" << endl;
}
// This code has been copied from bp.cpp, except where comments indicate TRWBP-specific behaviour
Real TRWBP::logZ() const {
Real sum = 0.0;
for( size_t I = 0; I < nrFactors(); I++ ) {
sum += (beliefF(I) * factor(I).log(true)).sum(); // TRWBP/FBP
sum += Weight(I) * beliefF(I).entropy(); // TRWBP/FBP
}
for( size_t i = 0; i < nrVars(); ++i ) {
Real c_i = 0.0;
bforeach( const Neighbor &I, nbV(i) )
c_i += Weight(I);
if( c_i != 1.0 )
sum += (1.0 - c_i) * beliefV(i).entropy(); // TRWBP/FBP
}
return sum;
}
std::vector<std::size_t> ExactInf::findMaximum() const {
Factor P;
for( size_t I = 0; I < nrFactors(); I++ )
P *= factor(I);
size_t linearState = P.p().argmax().first;
// convert to state
map<Var, size_t> state = calcState( P.vars(), linearState );
// convert to desired output data structure
vector<size_t> mapState;
mapState.reserve( nrVars() );
for( size_t i = 0; i < nrVars(); i++ )
mapState.push_back( state[var(i)] );
return mapState;
}
Factor ExactInf::calcMarginal( const VarSet &vs ) const {
Factor P;
for( size_t I = 0; I < nrFactors(); I++ )
P *= factor(I);
return P.marginal( vs, true );
}
void ExactInf::init() {
for( size_t i = 0; i < nrVars(); i++ )
_beliefsV[i].fill( 1.0 );
for( size_t I = 0; I < nrFactors(); I++ )
_beliefsF[I].fill( 1.0 );
}
