matthiasm@8: function bnet = learn_params_dbn(bnet, data)
matthiasm@8: % LEARN_PARAM_DBN Estimate params of a DBN for a fully observed model
matthiasm@8: % bnet = learn_params_dbn(bnet, data)
matthiasm@8: %
matthiasm@8: % data(i,t) is the value of node i in slice t (can be a cell array)
matthiasm@8: % We currently assume there is a single time series
matthiasm@8: %
matthiasm@8: % We set bnet.CPD{i} to its ML/MAP estimate.
matthiasm@8: %
matthiasm@8: % Currently we assume each node in the first 2 slices has its own CPD (no param tying);
matthiasm@8: % all nodes in slices >2 share their params with slice 2 as usual.
matthiasm@8: 
matthiasm@8: [ss T] = size(data);
matthiasm@8: 
matthiasm@8: % slice 1
matthiasm@8: for j=1:ss
matthiasm@8:   if adjustable_CPD(bnet.CPD{j})
matthiasm@8:     fam = family(bnet.dag,j);
matthiasm@8:     bnet.CPD{j} = learn_params(bnet.CPD{j}, data(fam,1));
matthiasm@8:   end
matthiasm@8: end
matthiasm@8: 
matthiasm@8: 
matthiasm@8: % slices 2:T
matthiasm@8: % data2(:,t) contains [data(:,t-1); data(:,t)].
matthiasm@8: % Then we extract out the rows corresponding to the parents in the current and previous slice.
matthiasm@8: data2 = [data(:,1:T-1);
matthiasm@8: 	 data(:,2:T)];
matthiasm@8: for j=1:ss
matthiasm@8:   j2 = j+ss;
matthiasm@8:   if adjustable_CPD(bnet.CPD{j2})
matthiasm@8:     fam = family(bnet.dag,j2);
matthiasm@8:     bnet.CPD{j2} = learn_params(bnet.CPD{j2}, data2(fam,:));
matthiasm@8:   end
matthiasm@8: end
matthiasm@8: