wolffd@0: function obslik = mk_hmm_obs_lik_vec(bnet, evidence)
wolffd@0: % MK_HMM_OBS_LIK_VEC Make the observation likelihood vector for one slice
wolffd@0: % obslik = mk_obs_lik(bnet, evidence)
wolffd@0: %
wolffd@0: % obslik(i) = Pr(y(t) | X(t)=i)
wolffd@0: % evidence{i,1} contains the evidence on node i in slice t-1
wolffd@0: % evidence{i,2} contains the evidence on node i in slice t
wolffd@0: 
wolffd@0: ns = bnet.node_sizes;
wolffd@0: ss = length(bnet.intra);
wolffd@0: onodes = find(~isemptycell(evidence(:)));
wolffd@0: hnodes = find(isemptycell(evidence(:)));
wolffd@0: ens = ns;
wolffd@0: ens(onodes) = 1;
wolffd@0: Q = prod(ens(hnodes));
wolffd@0: obslik = zeros(1,Q);
wolffd@0: dom = (1:ss)+ss;
wolffd@0: bigpot = dpot(dom, ens(dom));
wolffd@0: onodes1 = find(~isemptycell(evidence(:,1)));
wolffd@0: for i=onodes1(:)'
wolffd@0:   e = bnet.equiv_class(i,2);
wolffd@0:   fam = family(bnet.dag, i, 2);
wolffd@0:   pot = convert_to_pot(bnet.CPD{e}, 'd', fam, evidence);
wolffd@0:   bigpot = multiply_by_pot(bigpot, pot);
wolffd@0: end
wolffd@0: m = pot_to_marginal(bigpot);
wolffd@0: obslik = m.T(:);