wolffd@0: function [engine, loglik] = enter_evidence(engine, evidence, varargin)
wolffd@0: % ENTER_EVIDENCE Add the specified evidence to the network (kalman)
wolffd@0: % [engine, loglik] = enter_evidence(engine, evidence, ...)
wolffd@0: %
wolffd@0: % evidence{i,t} = [] if if X(i,t) is hidden, and otherwise contains its observed value (scalar or column vector)
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 - if 1, does max-product (same as sum-product for Gaussians!), else sum-product [0]
wolffd@0: % filter -   if 1, do filtering, else smoothing [0]
wolffd@0: %
wolffd@0: % e.g., engine = enter_evidence(engine, ev, 'maximize', 1)
wolffd@0: 
wolffd@0: maximize = 0;
wolffd@0: filter = 0;
wolffd@0: 
wolffd@0: % parse optional params
wolffd@0: args = varargin;
wolffd@0: nargs = length(args);
wolffd@0: if nargs > 0
wolffd@0:   for i=1:2:nargs
wolffd@0:     switch args{i},
wolffd@0:      case 'maximize', maximize = args{i+1}; 
wolffd@0:      case 'filter', filter = 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: assert(~maximize);
wolffd@0: 
wolffd@0: bnet = bnet_from_engine(engine);
wolffd@0: n = length(bnet.intra);
wolffd@0: onodes = bnet.observed;
wolffd@0: hnodes = mysetdiff(1:n, onodes);
wolffd@0: T = size(evidence, 2);
wolffd@0: ns = bnet.node_sizes;
wolffd@0: O = sum(ns(onodes));
wolffd@0: data = reshape(cat(1, evidence{onodes,:}), [O T]);
wolffd@0: 
wolffd@0: A = engine.trans_mat;
wolffd@0: C = engine.obs_mat;
wolffd@0: Q = engine.trans_cov;
wolffd@0: R = engine.obs_cov;
wolffd@0: init_x = engine.init_state;
wolffd@0: init_V = engine.init_cov;
wolffd@0: 
wolffd@0: if filter
wolffd@0:   [x, V, VV, loglik] = kalman_filter(data, A, C, Q, R, init_x, init_V);
wolffd@0: else
wolffd@0:   [x, V, VV, loglik] = kalman_smoother(data, A, C, Q, R, init_x, init_V);
wolffd@0: end
wolffd@0: 
wolffd@0:   
wolffd@0: % Wrap the posterior inside a potential, so it can be marginalized easily
wolffd@0: engine.one_slice_marginal = cell(1,T);
wolffd@0: engine.two_slice_marginal = cell(1,T);
wolffd@0: ns(onodes) = 0;
wolffd@0: ns(onodes+n) = 0;
wolffd@0: ss = length(bnet.intra);
wolffd@0: for t=1:T
wolffd@0:   dom = (1:n);
wolffd@0:   engine.one_slice_marginal{t} = mpot(dom+(t-1)*ss, ns(dom), 1, x(:,t), V(:,:,t));
wolffd@0: end
wolffd@0: % for t=1:T-1
wolffd@0: %   dom = (1:(2*n));
wolffd@0: %   mu = [x(:,t); x(:,t)];
wolffd@0: %   Sigma = [V(:,:,t) VV(:,:,t+1)';
wolffd@0: % 	   VV(:,:,t+1) V(:,:,t+1)];
wolffd@0: %   engine.two_slice_marginal{t} = mpot(dom+(t-1)*ss, ns(dom), 1, mu, Sigma);
wolffd@0: % end
wolffd@0: for t=2:T
wolffd@0:   %dom = (1:(2*n));
wolffd@0:   current_slice = hnodes;
wolffd@0:   next_slice = hnodes + ss;
wolffd@0:   dom = [current_slice next_slice];   
wolffd@0:   mu = [x(:,t-1); x(:,t)];
wolffd@0:   Sigma = [V(:,:,t-1) VV(:,:,t)';
wolffd@0: 	   VV(:,:,t) V(:,:,t)];
wolffd@0:   engine.two_slice_marginal{t-1} = mpot(dom+(t-2)*ss, ns(dom), 1, mu, Sigma);
wolffd@0: end