camir-aes2014: toolboxes/FullBNT-1.0.7/bnt/general/sample

first hg version after svn

rev	line source
wolffd@0	1 function sample = sample_bnet(bnet, varargin)
wolffd@0	2 % SAMPLE_BNET Generate a random sample from a Bayes net.
wolffd@0	3 % SAMPLE = SAMPLE_BNET(BNET, ...)
wolffd@0	4 %
wolffd@0	5 % sample{i} contains the value of the i'th node.
wolffd@0	6 % i.e., the result is an Nx1 cell array.
wolffd@0	7 % Nodes are sampled in the order given by bnet.order.
wolffd@0	8 %
wolffd@0	9 % Optional arguments:
wolffd@0	10 %
wolffd@0	11 % evidence - initial evidence; if evidence{i} is non-empty, node i won't be sampled.
wolffd@0	12
wolffd@0	13 % set defauly params
wolffd@0	14 n = length(bnet.dag);
wolffd@0	15 sample = cell(n,1);
wolffd@0	16
wolffd@0	17 % get optional params
wolffd@0	18 args = varargin;
wolffd@0	19 nargs = length(args);
wolffd@0	20 for i=1:2:nargs
wolffd@0	21 switch args{i},
wolffd@0	22 case 'evidence', sample = args{i+1}(:);
wolffd@0	23 otherwise, error(['unrecognized argument ' args{i}])
wolffd@0	24 end
wolffd@0	25 end
wolffd@0	26
wolffd@0	27 for j=bnet.order(:)'
wolffd@0	28 if isempty(sample{j})
wolffd@0	29 %ps = parents(bnet.dag, j);
wolffd@0	30 ps = bnet.parents{j};
wolffd@0	31 e = bnet.equiv_class(j);
wolffd@0	32 sample{j} = sample_node(bnet.CPD{e}, sample(ps));
wolffd@0	33 end
wolffd@0	34 end

Mercurial > hg > camir-aes2014