diff toolboxes/FullBNT-1.0.7/bnt/examples/static/StructLearn/model_select2.m @ 0:e9a9cd732c1e tip

first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
line wrap: on
line diff
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/toolboxes/FullBNT-1.0.7/bnt/examples/static/StructLearn/model_select2.m	Tue Feb 10 15:05:51 2015 +0000
@@ -0,0 +1,83 @@
+% Online Bayesian model selection demo.
+
+% We generate data from the model A->B
+% and compute the posterior prob of all 3 dags on 2 nodes:
+%  (1) A B,  (2) A <- B , (3) A -> B
+% Models 2 and 3 are Markov equivalent, and therefore indistinguishable from 
+% observational data alone.
+
+% We control the dependence of B on A by setting
+% P(B|A) = 0.5 - epislon and vary epsilon
+% as in Koller & Friedman book p512
+
+% ground truth
+N = 2;
+dag = zeros(N);
+A = 1; B = 2; 
+dag(A,B) = 1;
+
+ntrials = 100;
+ns = 2*ones(1,N);
+true_bnet = mk_bnet(dag, ns);
+true_bnet.CPD{1} = tabular_CPD(true_bnet, 1, [0.5 0.5]);
+
+% hypothesis space
+G = mk_all_dags(N);
+nhyp = length(G);
+hyp_bnet = cell(1, nhyp);
+for h=1:nhyp
+  hyp_bnet{h} = mk_bnet(G{h}, ns);
+  for i=1:N
+    % We must set the CPTs to the mean of the prior for sequential log_marg_lik to be correct
+    % The BDeu prior is score equivalent, so models 2,3 will be indistinguishable.
+    % The uniform Dirichlet prior is not score equivalent...
+    fam = family(G{h}, i);
+    hyp_bnet{h}.CPD{i}= tabular_CPD(hyp_bnet{h}, i, 'prior_type', 'dirichlet', ...
+				    'CPT', 'unif');
+  end
+end
+
+clf
+seeds = 1:3;
+expt = 1;
+for seedi=1:length(seeds)
+  seed = seeds(seedi);
+  rand('state', seed);
+  randn('state', seed);
+    
+  es = [0.05 0.1 0.15 0.2];
+  for ei=1:length(es)
+    e = es(ei);
+    true_bnet.CPD{2} = tabular_CPD(true_bnet, 2, [0.5+e 0.5-e; 0.5-e 0.5+e]);
+
+    prior = normalise(ones(1, nhyp));
+    hyp_w = zeros(ntrials+1, nhyp);
+    hyp_w(1,:) = prior(:)';
+    LL = zeros(1, nhyp);
+    ll = zeros(1, nhyp);
+    for t=1:ntrials
+      ev = cell2num(sample_bnet(true_bnet));
+      for i=1:nhyp
+	ll(i) = log_marg_lik_complete(hyp_bnet{i}, ev);
+	hyp_bnet{i} = bayes_update_params(hyp_bnet{i}, ev);
+      end
+      prior = normalise(prior .* exp(ll));
+      LL = LL + ll;
+      hyp_w(t+1,:) = prior;
+    end
+
+    % Plot posterior model probabilities
+    % Red = model 1 (no arcs), blue/green = models 2/3 (1 arc)
+    % Blue = model 2 (2->1)
+    % Green = model 3 (1->2, "ground truth")
+    
+    subplot2(length(seeds), length(es), seedi, ei);
+    m = size(hyp_w,1);
+    h=plot(1:m, hyp_w(:,1), 'r-',  1:m, hyp_w(:,2), 'b-.', 1:m, hyp_w(:,3), 'g:');
+    axis([0 m   0 1])
+    %title('model posterior vs. time')
+    title(sprintf('e=%3.2f, seed=%d', e, seed));
+    drawnow
+    expt = expt + 1;
+  end
+end