|
wolffd@0
|
1 function [data, clamped] = mk_mutilated_samples(bnet, ncases, max_clamp, usecell)
|
|
wolffd@0
|
2 % GEN_MUTILATED_SAMPLES Do random interventions and then draw random samples
|
|
wolffd@0
|
3 % [data, clamped] = gen_mutilated_samples(bnet, ncases, max_clamp, usecell)
|
|
wolffd@0
|
4 %
|
|
wolffd@0
|
5 % At each step, we pick a random subset of size 0 .. max_clamp, and
|
|
wolffd@0
|
6 % clamp these nodes to random values.
|
|
wolffd@0
|
7 %
|
|
wolffd@0
|
8 % data(i,m) is the value of node i in case m.
|
|
wolffd@0
|
9 % clamped(i,m) = 1 if node i in case m was set by intervention.
|
|
wolffd@0
|
10
|
|
wolffd@0
|
11 if nargin < 4, usecell = 1; end
|
|
wolffd@0
|
12
|
|
wolffd@0
|
13 ns = bnet.node_sizes;
|
|
wolffd@0
|
14 n = length(bnet.dag);
|
|
wolffd@0
|
15 if usecell
|
|
wolffd@0
|
16 data = cell(n, ncases);
|
|
wolffd@0
|
17 else
|
|
wolffd@0
|
18 data = zeros(n, ncases);
|
|
wolffd@0
|
19 end
|
|
wolffd@0
|
20 clamped = zeros(n, ncases);
|
|
wolffd@0
|
21
|
|
wolffd@0
|
22 csubsets = subsets(1:n, max_clamp, 0); % includes the empty set
|
|
wolffd@0
|
23 distrib_cset = normalise(ones(1, length(csubsets)));
|
|
wolffd@0
|
24
|
|
wolffd@0
|
25 for m=1:ncases
|
|
wolffd@0
|
26 cset = csubsets{sample_discrete(distrib_cset)};
|
|
wolffd@0
|
27 nvals = prod(ns(cset));
|
|
wolffd@0
|
28 distrib_cvals = normalise(ones(1, nvals));
|
|
wolffd@0
|
29 cvals = ind2subv(ns(cset), sample_discrete(distrib_cvals));
|
|
wolffd@0
|
30 mutilated_bnet = do_intervention(bnet, cset, cvals);
|
|
wolffd@0
|
31 ev = sample_bnet(mutilated_bnet);
|
|
wolffd@0
|
32 if usecell
|
|
wolffd@0
|
33 data(:,m) = ev;
|
|
wolffd@0
|
34 else
|
|
wolffd@0
|
35 data(:,m) = cell2num(ev);
|
|
wolffd@0
|
36 end
|
|
wolffd@0
|
37 clamped(cset,m) = 1;
|
|
wolffd@0
|
38 end
|
