wolffd@0: function [marginal, msg, loglik] = filter_evidence_old(engine, evidence)
wolffd@0: % [marginal, msg, loglik] = filter_evidence(engine, evidence) (pearl_dbn)
wolffd@0: 
wolffd@0: [ss T] = size(evidence);
wolffd@0: bnet = bnet_from_engine(engine);
wolffd@0: bnet2 = dbn_to_bnet(bnet, T);
wolffd@0: ns = bnet2.node_sizes;
wolffd@0: hnodes = mysetdiff(1:ss, engine.onodes);
wolffd@0: hnodes = hnodes(:)';
wolffd@0: 
wolffd@0: [engine.parent_index, engine.child_index] = mk_pearl_msg_indices(bnet2);
wolffd@0: 
wolffd@0: msg = init_msgs(bnet2.dag, ns, evidence);
wolffd@0: msg = init_ev_msgs(engine, evidence, msg);
wolffd@0: 
wolffd@0: verbose = 1;
wolffd@0: if verbose, fprintf('\nold filtering\n'); end
wolffd@0: 
wolffd@0: for t=1:T
wolffd@0:   % update pi
wolffd@0:   for i=hnodes
wolffd@0:     n = i + (t-1)*ss;
wolffd@0:     ps = parents(bnet2.dag, n);
wolffd@0:     if t==1
wolffd@0:       e = bnet.equiv_class(i,1);
wolffd@0:     else
wolffd@0:       e = bnet.equiv_class(i,2);
wolffd@0:     end
wolffd@0:     msg{n}.pi = compute_pi(bnet.CPD{e}, n, ps, msg);
wolffd@0:     %if verbose, fprintf('%d computes pi\n', n); disp(msg{n}.pi); end
wolffd@0:     msg{n}.pi = normalise(msg{n}.pi(:) .* msg{n}.lambda_from_self(:));
wolffd@0:     if verbose, fprintf('%d recomputes pi\n', n); disp(msg{n}.pi); end
wolffd@0:   end
wolffd@0:   % send pi msg to children
wolffd@0:   for i=hnodes
wolffd@0:     n = i + (t-1)*ss;
wolffd@0:     cs = children(bnet2.dag, n);
wolffd@0:     for c=cs(:)'
wolffd@0:       j = engine.parent_index{c}(n); % n is c's j'th parent
wolffd@0:       pi_msg = normalise(compute_pi_msg(n, cs, msg, c, ns));
wolffd@0:       msg{c}.pi_from_parent{j} = pi_msg;
wolffd@0:       if verbose, fprintf('%d sends pi to %d\n', n,c); disp(pi_msg); end
wolffd@0:     end
wolffd@0:   end
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: marginal = cell(ss,T);
wolffd@0: lik = zeros(1,ss*T);
wolffd@0: for t=1:T
wolffd@0:   for i=1:ss
wolffd@0:     n = i + (t-1)*ss;
wolffd@0:     %[bel, lik(n)] = normalise(msg{n}.pi .* msg{n}.lambda);     
wolffd@0:     [bel, lik(n)] = normalise(msg{n}.pi);
wolffd@0:     marginal{i,t} = bel;
wolffd@0:   end
wolffd@0: end
wolffd@0: 
wolffd@0: loglik = sum(log(lik));
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: %%%%%%%
wolffd@0: 
wolffd@0: function lambda = compute_lambda(n, cs, msg, ns)
wolffd@0: % Pearl p183 eq 4.50
wolffd@0: lambda = prod_lambda_msgs(n, cs, msg, ns);
wolffd@0: 
wolffd@0: %%%%%%%
wolffd@0: 
wolffd@0: function pi_msg = compute_pi_msg(n, cs, msg, c, ns)
wolffd@0: % Pearl p183 eq 4.53 and 4.51
wolffd@0: pi_msg = msg{n}.pi .* prod_lambda_msgs(n, cs, msg, ns, c);
wolffd@0: 
wolffd@0: %%%%%%%%%
wolffd@0: 
wolffd@0: function lam = prod_lambda_msgs(n, cs, msg, ns, except)
wolffd@0: 
wolffd@0: if nargin < 5, except = -1; end
wolffd@0: 
wolffd@0: %lam = msg{n}.lambda_from_self(:);
wolffd@0: lam = ones(ns(n), 1);
wolffd@0: for i=1:length(cs)
wolffd@0:   c = cs(i);
wolffd@0:   if c ~= except
wolffd@0:     lam = lam .* msg{n}.lambda_from_child{i};
wolffd@0:   end
wolffd@0: end   
wolffd@0: 
wolffd@0: 
wolffd@0: %%%%%%%%%%%
wolffd@0: 
wolffd@0: function msg = init_msgs(dag, ns, evidence)
wolffd@0: % INIT_MSGS Initialize the lambda/pi message and state vectors (pearl_dbn)
wolffd@0: % msg =  init_msgs(dag, ns, evidence)
wolffd@0: %
wolffd@0: % We assume all the hidden nodes are discrete.
wolffd@0: 
wolffd@0: N = length(dag);
wolffd@0: msg = cell(1,N);
wolffd@0: observed = ~isemptycell(evidence(:));
wolffd@0: 
wolffd@0: for n=1:N
wolffd@0:   ps = parents(dag, n);
wolffd@0:   msg{n}.pi_from_parent = cell(1, length(ps));
wolffd@0:   for i=1:length(ps)
wolffd@0:     p = ps(i);
wolffd@0:     msg{n}.pi_from_parent{i} = ones(ns(p), 1);
wolffd@0:   end
wolffd@0:   
wolffd@0:   cs = children(dag, n);
wolffd@0:   msg{n}.lambda_from_child = cell(1, length(cs));
wolffd@0:   for i=1:length(cs)
wolffd@0:     c = cs(i);
wolffd@0:     msg{n}.lambda_from_child{i} = ones(ns(n), 1);
wolffd@0:   end
wolffd@0: 
wolffd@0:   msg{n}.lambda = ones(ns(n), 1);
wolffd@0:   msg{n}.pi = ones(ns(n), 1);
wolffd@0:   
wolffd@0:   msg{n}.lambda_from_self = ones(ns(n), 1);
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: %%%%%%%%%
wolffd@0: 
wolffd@0: function msg = init_ev_msgs(engine, evidence, msg)
wolffd@0: % Initialize the lambdas with any evidence
wolffd@0: 
wolffd@0: [ss T] = size(evidence);
wolffd@0: bnet = bnet_from_engine(engine);
wolffd@0: pot_type = 'd';
wolffd@0: t = 1;
wolffd@0: hnodes = mysetdiff(1:ss, engine.onodes);
wolffd@0: for i=hnodes(:)'
wolffd@0:   c = engine.obschild(i);
wolffd@0:   if c > 0
wolffd@0:     fam = family(bnet.dag, c);
wolffd@0:     e = bnet.equiv_class(c, 1);
wolffd@0:     CPDpot = CPD_to_pot(pot_type, bnet.CPD{e}, fam, bnet.node_sizes(:), bnet.cnodes(:), evidence(:,1));
wolffd@0:     temp = pot_to_marginal(CPDpot);
wolffd@0:     n = i;
wolffd@0:     msg{n}.lambda_from_self = temp.T;
wolffd@0:   end
wolffd@0: end
wolffd@0: for t=2:T
wolffd@0:   for i=hnodes(:)'
wolffd@0:     c = engine.obschild(i);
wolffd@0:     if c > 0 
wolffd@0:       fam = family(bnet.dag, c, 2);
wolffd@0:       e = bnet.equiv_class(c, 2);
wolffd@0:       CPDpot = CPD_to_pot(pot_type, bnet.CPD{e}, fam, bnet.node_sizes(:), bnet.cnodes(:), evidence(:,t-1:t));
wolffd@0:       temp = pot_to_marginal(CPDpot);
wolffd@0:       n = i + (t-1)*ss;
wolffd@0:       msg{n}.lambda_from_self = temp.T;
wolffd@0:     end
wolffd@0:   end
wolffd@0: end