camir-aes2014: toolboxes/FullBNT-1.0.7/bnt/examples/static/StructLearn/mcmc1.m annotate

first hg version after svn

rev	line source
wolffd@0	1 % We compare MCMC structure learning with exhaustive enumeration of all dags.
wolffd@0	2
wolffd@0	3 N = 3;
wolffd@0	4 %N = 4;
wolffd@0	5 dag = mk_rnd_dag(N);
wolffd@0	6 ns = 2*ones(1,N);
wolffd@0	7 bnet = mk_bnet(dag, ns);
wolffd@0	8 for i=1:N
wolffd@0	9 bnet.CPD{i} = tabular_CPD(bnet, i);
wolffd@0	10 end
wolffd@0	11
wolffd@0	12 ncases = 100;
wolffd@0	13 data = zeros(N, ncases);
wolffd@0	14 for m=1:ncases
wolffd@0	15 data(:,m) = cell2num(sample_bnet(bnet));
wolffd@0	16 end
wolffd@0	17
wolffd@0	18 dags = mk_all_dags(N);
wolffd@0	19 score = score_dags(data, ns, dags);
wolffd@0	20 post = normalise(exp(score));
wolffd@0	21
wolffd@0	22 [sampled_graphs, accept_ratio] = learn_struct_mcmc(data, ns, 'nsamples', 100, 'burnin', 10);
wolffd@0	23 mcmc_post = mcmc_sample_to_hist(sampled_graphs, dags);
wolffd@0	24
wolffd@0	25 if 0
wolffd@0	26 subplot(2,1,1)
wolffd@0	27 bar(post)
wolffd@0	28 subplot(2,1,2)
wolffd@0	29 bar(mcmc_post)
wolffd@0	30 print(gcf, '-djpeg', '/home/cs/murphyk/public_html/Bayes/Figures/mcmc_post.jpg')
wolffd@0	31
wolffd@0	32 clf
wolffd@0	33 plot(accept_ratio)
wolffd@0	34 print(gcf, '-djpeg', '/home/cs/murphyk/public_html/Bayes/Figures/mcmc_accept.jpg')
wolffd@0	35 end

Mercurial > hg > camir-aes2014