wolffd@0: function CPD = learn_params(CPD, fam, data, ns, cnodes)
wolffd@0: %function CPD = learn_params(CPD, fam, data, ns, cnodes)
wolffd@0: % LEARN_PARAMS Compute the maximum likelihood estimate of the params of a gaussian CPD given complete data
wolffd@0: % CPD = learn_params(CPD, fam, data, ns, cnodes)
wolffd@0: %
wolffd@0: % data(i,m) is the value of node i in case m (can be cell array).
wolffd@0: % We assume this node has a maximize_params method.
wolffd@0: 
wolffd@0: ncases = size(data, 2);
wolffd@0: CPD = reset_ess(CPD);
wolffd@0: % make a fully observed joint distribution over the family
wolffd@0: fmarginal.domain = fam;
wolffd@0: fmarginal.T = 1;
wolffd@0: fmarginal.mu = [];
wolffd@0: fmarginal.Sigma = [];
wolffd@0: if ~iscell(data)
wolffd@0:   cases = num2cell(data);
wolffd@0: else
wolffd@0:   cases = data;
wolffd@0: end
wolffd@0: hidden_bitv = zeros(1, max(fam));
wolffd@0: for m=1:ncases
wolffd@0:   % specify (as a bit vector) which elements in the family domain are hidden
wolffd@0:   hidden_bitv = zeros(1, max(fmarginal.domain));
wolffd@0:   ev = cases(:,m);
wolffd@0:   hidden_bitv(find(isempty(ev)))=1;
wolffd@0:   CPD = update_ess(CPD, fmarginal, ev, ns, cnodes, hidden_bitv);
wolffd@0: end
wolffd@0: CPD = maximize_params(CPD);
wolffd@0: 
wolffd@0: