wolffd@0: function [engine, ll, niter] = enter_evidence(engine, evidence, varargin)
wolffd@0: % ENTER_EVIDENCE Propagate evidence using belief propagation
wolffd@0: % [engine, ll, niter] = enter_evidence(engine, evidence, ...)
wolffd@0: %
wolffd@0: % The log-likelihood is not computed; ll = 0.
wolffd@0: % niter contains the number of iterations used 
wolffd@0: %
wolffd@0: % The following optional arguments can be specified in the form of name/value pairs:
wolffd@0: % [default value in brackets]
wolffd@0: %
wolffd@0: % maximize - 1 means use max-product, 0 means use sum-product [0]
wolffd@0: %
wolffd@0: % e.g., engine = enter_evidence(engine, ev, 'maximize', 1)
wolffd@0: 
wolffd@0: ll = 0;
wolffd@0: maximize = 0;
wolffd@0: 
wolffd@0: if nargin >= 3
wolffd@0:   args = varargin;
wolffd@0:   nargs = length(args);
wolffd@0:   for i=1:2:nargs
wolffd@0:     switch args{i},
wolffd@0:      case 'maximize', maximize = args{i+1};
wolffd@0:      otherwise,
wolffd@0:       error(['invalid argument name ' args{i}]);
wolffd@0:     end
wolffd@0:   end
wolffd@0: end
wolffd@0: 
wolffd@0: verbose = 0;
wolffd@0: 
wolffd@0: ns = engine.fgraph.node_sizes;
wolffd@0: onodes = find(~isemptycell(evidence));
wolffd@0: hnodes = find(isemptycell(evidence));
wolffd@0: cnodes = engine.fgraph.cnodes;
wolffd@0: pot_type = determine_pot_type(engine.fgraph, onodes);
wolffd@0: 
wolffd@0: % prime each local kernel with evidence (if any)
wolffd@0: nfactors = engine.fgraph.nfactors;
wolffd@0: nvars = engine.fgraph.nvars;
wolffd@0: factors = cell(1,nfactors);
wolffd@0: for f=1:nfactors
wolffd@0:   K = engine.fgraph.factors{engine.fgraph.equiv_class(f)};
wolffd@0:   factors{f} = convert_to_pot(K, pot_type, engine.fgraph.dom{f}(:), evidence);
wolffd@0: end
wolffd@0:   
wolffd@0: % initialise msgs
wolffd@0: msg_var_to_fac = cell(nvars, nfactors);
wolffd@0: for x=1:nvars
wolffd@0:   for f=engine.fgraph.dep{x}
wolffd@0:     msg_var_to_fac{x,f} = mk_initial_pot(pot_type, x, ns, cnodes, onodes);
wolffd@0:   end
wolffd@0: end
wolffd@0: msg_fac_to_var = cell(nfactors, nvars);
wolffd@0: dom = cell(1, nfactors);
wolffd@0: for f=1:nfactors
wolffd@0:   %hdom{f} = myintersect(engine.fgraph.dom{f}, hnodes);
wolffd@0:   dom{f} = engine.fgraph.dom{f}(:)';
wolffd@0:   for x=dom{f}
wolffd@0:     msg_fac_to_var{f,x} = mk_initial_pot(pot_type, x, ns, cnodes, onodes);
wolffd@0:     %msg_fac_to_var{f,x} = marginalize_pot(factors{f}, x);
wolffd@0:   end
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: converged = 0;
wolffd@0: iter = 1;
wolffd@0: var_prod = cell(1, nvars);
wolffd@0: fac_prod = cell(1, nfactors);
wolffd@0: 
wolffd@0: while ~converged & (iter <= engine.max_iter)
wolffd@0:   if verbose, fprintf('iter %d\n', iter);  end
wolffd@0:   
wolffd@0:   % absorb
wolffd@0:   old_var_prod = var_prod;
wolffd@0:   for x=1:nvars
wolffd@0:     var_prod{x} = mk_initial_pot(pot_type, x, ns, cnodes, onodes);
wolffd@0:     for f=engine.fgraph.dep{x}
wolffd@0:       var_prod{x} = multiply_by_pot(var_prod{x}, msg_fac_to_var{f,x});
wolffd@0:     end
wolffd@0:   end
wolffd@0:   for f=1:nfactors
wolffd@0:     fac_prod{f} = mk_initial_pot(pot_type, dom{f}, ns, cnodes, onodes);
wolffd@0:     for x=dom{f}
wolffd@0:       fac_prod{f} = multiply_by_pot(fac_prod{f}, msg_var_to_fac{x,f});
wolffd@0:     end
wolffd@0:   end
wolffd@0: 
wolffd@0:   % send msgs to neighbors
wolffd@0:   old_msg_var_to_fac = msg_var_to_fac;
wolffd@0:   old_msg_fac_to_var = msg_fac_to_var;
wolffd@0:   converged = 1;
wolffd@0:   for x=1:nvars
wolffd@0:     %if verbose, disp(['var ' num2str(x) ' sending to fac ' num2str(engine.fgraph.dep{x})]); end
wolffd@0:     for f=engine.fgraph.dep{x}
wolffd@0:       temp = divide_by_pot(var_prod{x}, old_msg_fac_to_var{f,x});
wolffd@0:       msg_var_to_fac{x,f} = normalize_pot(temp);
wolffd@0:       if ~approxeq_pot(msg_var_to_fac{x,f}, old_msg_var_to_fac{x,f}, engine.tol), converged = 0; end
wolffd@0:     end
wolffd@0:   end
wolffd@0:   for f=1:nfactors
wolffd@0:     %if verbose, disp(['fac ' num2str(f) ' sending to var ' num2str(dom{f})]); end
wolffd@0:     for x=dom{f}
wolffd@0:       temp = divide_by_pot(fac_prod{f}, old_msg_var_to_fac{x,f});
wolffd@0:       temp2 = multiply_by_pot(factors{f}, temp);
wolffd@0:       temp3 = marginalize_pot(temp2, x, maximize);
wolffd@0:       msg_fac_to_var{f,x} = normalize_pot(temp3);
wolffd@0:       if ~approxeq_pot(msg_fac_to_var{f,x}, old_msg_fac_to_var{f,x}, engine.tol), converged = 0; end
wolffd@0:     end
wolffd@0:   end
wolffd@0: 
wolffd@0:   if iter==1
wolffd@0:     converged = 0;
wolffd@0:   end
wolffd@0:   iter = iter + 1;
wolffd@0: end
wolffd@0: 
wolffd@0: niter = iter - 1;
wolffd@0: engine.niter = niter;
wolffd@0: 
wolffd@0: for x=1:nvars
wolffd@0:   engine.marginal_nodes{x} = normalize_pot(var_prod{x});
wolffd@0: end
wolffd@0: 
wolffd@0: