wolffd@0: function L = log_nextcase_prob_node(CPD, self_ev, pev, test_self_ev, test_pev)
wolffd@0: % LOG_NEXTCASE_PROB_NODE compute the joint distribution of a node (tabular) of a new case given
wolffd@0: % completely observed data.
wolffd@0: %
wolffd@0: % The input arguments are mainly similar with log_marg_prob_node(CPD, self_ev, pev, usecell),
wolffd@0: % but add test_self_ev, test_pev, and without usecell
wolffd@0: % test_self_ev(m) is the evidence on this node in a test case.
wolffd@0: % test_pev(i) is the evidence on the i'th parent in the test case (if there are any parents).
wolffd@0: %
wolffd@0: % Written by qian.diao@intel.com
wolffd@0: 
wolffd@0: ncases = length(self_ev);
wolffd@0: sz = CPD.sizes;
wolffd@0: nparents = length(sz)-1;
wolffd@0: assert(ncases == size(pev, 2)); 
wolffd@0: 
wolffd@0: if nargin < 6
wolffd@0:   %usecell = 0;
wolffd@0:   if iscell(self_ev)
wolffd@0:     usecell = 1;
wolffd@0:   else
wolffd@0:     usecell = 0;
wolffd@0:   end
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: if ncases==0
wolffd@0:   L = 0;
wolffd@0:   return;
wolffd@0: elseif ncases==1  % speedup the sequential learning case; here need correction!!!
wolffd@0:   CPT = CPD.CPT;
wolffd@0:   % We assume the CPTs are already set to the mean of the posterior (due to bayes_update_params)
wolffd@0:   if usecell
wolffd@0:     x = cat(1, pev{:})';
wolffd@0:     y = self_ev{1};
wolffd@0:   else
wolffd@0:     %x = pev(:)';
wolffd@0:     x = pev;
wolffd@0:     y = self_ev;
wolffd@0:   end
wolffd@0:   switch nparents
wolffd@0:    case 0, p = CPT(y);
wolffd@0:    case 1, p = CPT(x(1), y);
wolffd@0:    case 2, p = CPT(x(1), x(2), y);
wolffd@0:    case 3, p = CPT(x(1), x(2), x(3), y);
wolffd@0:    otherwise,
wolffd@0:     ind = subv2ind(sz, [x y]);
wolffd@0:     p = CPT(ind);
wolffd@0:   end
wolffd@0:   L = log(p);
wolffd@0: else
wolffd@0:   % We ignore the CPTs here and assume the prior has not been changed
wolffd@0:   
wolffd@0:   % We arrange the data as in the following example.
wolffd@0:   % Let there be 2 parents and 3 cases. Let p(i,m) be parent i in case m,
wolffd@0:   % and y(m) be the child in case m. Then we create the data matrix
wolffd@0:   % 
wolffd@0:   % p(1,1) p(1,2) p(1,3)
wolffd@0:   % p(2,1) p(2,2) p(2,3)
wolffd@0:   % y(1)   y(2)   y(3)
wolffd@0:   if usecell
wolffd@0:     data = [cell2num(pev); cell2num(self_ev)]; 
wolffd@0:   else
wolffd@0:     data = [pev; self_ev];
wolffd@0:   end
wolffd@0:   counts = compute_counts(data, sz);
wolffd@0:   
wolffd@0:   % compute the (N_ijk'+ N_ijk)/(N_ij' + N_ij) under the condition of 1_m+1,ijk = 1 
wolffd@0:   L = predict_family(counts, CPD.prior, test_self_ev, test_pev);
wolffd@0: end
wolffd@0: 
wolffd@0: