Mercurial > hg > camir-aes2014
diff toolboxes/FullBNT-1.0.7/bnt/examples/static/Models/mk_cancer_bnet.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/Models/mk_cancer_bnet.m Tue Feb 10 15:05:51 2015 +0000 @@ -0,0 +1,61 @@ +function bnet = mk_cancer_bnet(CPD_type, p) +% MK_CANCER_BNET Make the 'Cancer' Bayes net. +% +% BNET = MK_CANCER_BNET uses the noisy-or parameters specified in Fig 4a of the UAI98 paper by +% Friedman, Murphy and Russell, "Learning the Structure of DPNs", p145. +% +% BNET = MK_CANCER_BNET('noisyor', p) makes each CPD a noisy-or, with probability p of +% suppression for each parent; leaks are turned off. +% +% BNET = MK_CANCER_BNET('cpt', p) uses random CPT parameters drawn from a Dirichlet(p,p,...) +% distribution. If p << 1, this is near deterministic; if p >> 1, this is near 1/k. +% p defaults to 1.0 (uniform distribution). +% +% BNET = MK_CANCER_BNET('bool') makes each CPT a random boolean function. +% +% In all cases, the root is set to a uniform distribution. + +if nargin == 0 + rnd = 0; +else + rnd = 1; +end + +n = 5; +dag = zeros(n); +dag(1,[2 3]) = 1; +dag(2,4) = 1; +dag(3,4) = 1; +dag(4,5) = 1; + +ns = 2*ones(1,n); +bnet = mk_bnet(dag, ns); + +if ~rnd + bnet.CPD{1} = tabular_CPD(bnet, 1, [0.5 0.5]); + bnet.CPD{2} = noisyor_CPD(bnet, 2, 1.0, 1-0.9); + bnet.CPD{3} = noisyor_CPD(bnet, 3, 1.0, 1-0.2); + bnet.CPD{4} = noisyor_CPD(bnet, 4, 1.0, 1-[0.7 0.6]); + bnet.CPD{5} = noisyor_CPD(bnet, 5, 1.0, 1-0.5); +else + switch CPD_type + case 'noisyor', + for i=1:n + ps = parents(dag, i); + bnet.CPD{i} = noisyor_CPD(bnet, i, 1.0, p*ones(1,length(ps))); + end + case 'bool', + for i=1:n + bnet.CPD{i} = boolean_CPD(bnet, i, 'rnd'); + end + case 'cpt', + for i=1:n + bnet.CPD{i} = tabular_CPD(bnet, i, p); + end + otherwise + error(['bad CPD type ' CPD_type]); + end +end + + +