Factor Inference::FactorMarginalization ( Factor A, vec ele)
{
//function B = FactorMarginalization(A,V)
// this function is only for hmm, ele has only one var
// var......
Factor B;
// eliminate te first var
B.var << A.var(1);
B.card << A.card(0);
mat Val = A.val.t();
Val = reshape(Val,B.card(0),B.card(0));
// eliminate the second var
if (ele(0) == A.var(1))
{
Val = Val.t();
B.var << A.var(0);
}
//B.val = log(sum(exp(bsxfun(@minus, Val, max(Val)))))+max(Val);
B.val = zeros<vec>(Val.n_cols);
for (int step = 0; step != Val.n_cols; step ++)
{
B.val(step) = Utils::logsumexp( Val.col(step) );
}
return B;
}
CliqueTree Inference::CreateCliqueTreeHMMFast( vector<Factor> factors )
{
int maxVar = 0;
// get the max var id
for ( vector<Factor>::iterator iter = factors.begin();
iter != factors.end(); iter ++ )
{
double max_now = iter->var.max();
if ( max_now > maxVar )
{
maxVar = max_now;
}
}
int numNodes = maxVar - 1;
int card = factors[0].card(0);
CliqueTree P(numNodes);
//P.cliqueList = repmat(struct('var', [], 'card', [], 'val', []), numNodes, 1);
//P.edges = zeros(numNodes);
for ( int i = 0; i != numNodes; i++ )
{
P.cliqueList[i].var << i+1 << i+2;
P.cliqueList[i].card << card << card;
P.cliqueList[i].val = ones<vec>(card * card);
if (i > 0)
{
P.edges(i, i-1) = 1;
P.edges(i-1, i) = 1;
}
}
// the name of the variable starts from 1 !!!!
for ( int i = 0; i != factors.size(); i ++ )
{
Factor f = factors[i];
int cliqueIdx = 0;
if (f.var.n_rows == 1)
{
if (f.var(0) > 1)
{
cliqueIdx = f.var(0) - 1;
}else{
cliqueIdx = 1;
}
vec updateIdxs;
mat assignments = zeros<mat>(card, 2);
vec cards;
cards << card << card;
for ( int assignment = 0; assignment != card; assignment ++ )
{
if (f.var(0) == cliqueIdx)
{
assignments.col(0) = linspace<vec>(0, card - 1, card);
assignments.col(1) = assignment*ones<vec>(card);
updateIdxs = Utils::AssignmentToIndex(assignments, cards);
}else{
assignments.col(1) = linspace<vec>(0, card - 1, card);
assignments.col(0) = assignment*ones<vec>(card);
updateIdxs = Utils::AssignmentToIndex(assignments, cards);
}
for (int step = 0; step != updateIdxs.n_rows; step ++)
{
P.cliqueList[cliqueIdx - 1].val(updateIdxs(step)) += f.val(step);
}
}
}else{
if ( f.var.n_rows != 2 )
{
std::cout << "ERROR: var more than 2!" << std::endl;
}
cliqueIdx = min(f.var);
if (f.var(0) > f.var(1))
{
// % sort the factor val so it's in increasing var order
uvec order = sort_index(f.var); //%#ok
mat oldAssignments = Utils::IndexToAssignment(linspace<vec>(0,f.val.n_rows-1,f.val.n_rows), f.card);
mat newAssignments = oldAssignments;
// (:, order)
for (int step = 0; step != oldAssignments.n_cols; step ++)
{
newAssignments.col(step) = oldAssignments.col(order(step));
}
vec new_card = f.card;
for (int step = 0; step != f.card.n_rows; step ++)
{
new_card(step) = f.card(order(step));
}
f.card = new_card;
f.var = sort(f.var);
vec new_index = Utils::AssignmentToIndex(newAssignments, f.card);
vec new_val = f.val;
for ( int step = 0; step != f.val.n_rows; step ++ )
{
new_val(step) = f.val( new_index(step) );
}
f.val = new_val;
}
P.cliqueList[cliqueIdx - 1].val += f.val;
}
}
return P;
}
