wolffd@0: % Construct various DBNs and examine their clique structure.
wolffd@0: % This was used to generate various figures in chap 3-4 of my thesis.
wolffd@0: 
wolffd@0: % Examine the cliques in the unrolled mildew net
wolffd@0: 
wolffd@0: %dbn = mk_mildew_dbn;
wolffd@0: dbn = mk_chmm(4);
wolffd@0: ss = dbn.nnodes_per_slice;
wolffd@0: T = 7;
wolffd@0: N = ss*T;
wolffd@0: bnet = dbn_to_bnet(dbn, T);
wolffd@0: 
wolffd@0: constrained = 0;
wolffd@0: if constrained
wolffd@0:   stages = num2cell(unroll_set(1:ss, ss, T), 1);
wolffd@0: else
wolffd@0:   stages = { 1:N; };
wolffd@0: end
wolffd@0: clusters = {};
wolffd@0: %[jtree, root, cliques, B, w, elim_order, moral_edges, fill_in_edges] = ...
wolffd@0: %    dag_to_jtree(bnet, bnet.observed, stages, clusters);
wolffd@0: [jtree, root, cliques] =  graph_to_jtree(moralize(bnet.dag), ones(1,N), stages, clusters);
wolffd@0: 
wolffd@0: flip=1;
wolffd@0: clf;[dummyx, dummyy, h] = draw_dbn(dbn.intra, dbn.inter, flip, T, -1);
wolffd@0: dir = '/home/eecs/murphyk/WP/Thesis/Figures/Inf/MildewUnrolled';
wolffd@0: mk_ps_from_clqs(dbn, T, cliques, [])
wolffd@0: %mk_collage_from_clqs(dir, cliques)
wolffd@0: 
wolffd@0: 
wolffd@0: % Examine the cliques in the cascade DBN
wolffd@0: 
wolffd@0: % A-A
wolffd@0: %  \
wolffd@0: % B B
wolffd@0: %  \
wolffd@0: % C C
wolffd@0: %  \
wolffd@0: % D D
wolffd@0: ss = 4;
wolffd@0: intra = zeros(ss);
wolffd@0: inter = zeros(ss);
wolffd@0: inter(1, [1 2])=1;
wolffd@0: for i=2:ss-1
wolffd@0:   inter(i,i+1)=1;
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: % 2 coupled HMMs 1,3  and 2,4
wolffd@0: ss = 4;
wolffd@0: intra = zeros(ss);
wolffd@0: inter = zeros(ss); % no persistent edges
wolffd@0: %inter = diag(ones(ss,1)); % persitence edges
wolffd@0: inter(1,3)=1; inter(3,1)=1;
wolffd@0: inter(2,4)=1; inter(4,2)=1;
wolffd@0: 
wolffd@0: %bnet = mk_fhmm(3);
wolffd@0: bnet = mk_chmm(4);
wolffd@0: intra = bnet.intra;
wolffd@0: inter = bnet.inter;
wolffd@0: 
wolffd@0: clqs = compute_minimal_interface(intra, inter);
wolffd@0: celldisp(clqs)
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: % A A
wolffd@0: %  \
wolffd@0: % B B
wolffd@0: %  \
wolffd@0: % C C
wolffd@0: %  \
wolffd@0: % D-D
wolffd@0: ss = 4;
wolffd@0: intra = zeros(ss);
wolffd@0: inter = zeros(ss);
wolffd@0: for i=1:ss-1
wolffd@0:   inter(i,i+1)=1;
wolffd@0: end
wolffd@0: inter(4,4)=1;
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: ns = 2*ones(1,ss);
wolffd@0: dbn = mk_dbn(intra, inter, ns);
wolffd@0: for i=2*ss
wolffd@0:   dbn.CPD{i} = tabular_CPD(bnet, i);
wolffd@0: end
wolffd@0: 
wolffd@0: T = 4;
wolffd@0: N = ss*T;
wolffd@0: bnet = dbn_to_bnet(dbn, T);
wolffd@0: 
wolffd@0: constrained = 1;
wolffd@0: if constrained
wolffd@0:   % elim first 3 slices first in any order
wolffd@0:   stages = {1:12, 13:16};
wolffd@0:   %stages = num2cell(unroll_set(1:ss, ss, T), 1);
wolffd@0: else
wolffd@0:   stages = { 1:N; };
wolffd@0: end
wolffd@0: clusters = {};
wolffd@0: %[jtree, root, cliques, B, w, elim_order, moral_edges, fill_in_edges] = ...
wolffd@0: %    dag_to_jtree(bnet, bnet.observed, stages, clusters);
wolffd@0: [jtree, root, cliques] =  graph_to_jtree(moralize(bnet.dag), ones(1,N), stages, clusters);
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: % Examine the cliques in the 1.5 slice DBN
wolffd@0: 
wolffd@0: %dbn = mk_mildew_dbn;
wolffd@0: dbn = mk_water_dbn;
wolffd@0: %dbn = mk_bat_dbn;
wolffd@0: ss = dbn.nnodes_per_slice;
wolffd@0: int = compute_fwd_interface(dbn);
wolffd@0: bnet15 = mk_slice_and_half_dbn(dbn, int);
wolffd@0: N = length(bnet15.dag);
wolffd@0: stages = {1:N};
wolffd@0: 
wolffd@0: % bat
wolffd@0: %cl1 = [16 17 19 7 14];
wolffd@0: %cl2 = [27 25 21 23 20];
wolffd@0: %clusters = {cl1, cl2, cl1+ss, cl2+ss};
wolffd@0: 
wolffd@0: % water
wolffd@0: %cl1 = 1:2; cl2 = 3:6; cl3 = 7:8;
wolffd@0: %clusters = {cl1, cl2, cl3, cl1+ss, cl2+ss, cl3+ss};
wolffd@0: 
wolffd@0: %clusters = {};
wolffd@0: clusters = {int, int+ss};
wolffd@0: %[jtree, root, cliques, B, w, elim_order, moral_edges, fill_in_edges] = ...
wolffd@0: %    dag_to_jtree(bnet15, bnet.observed, stages, clusters);
wolffd@0: [jtree, root, cliques] =  graph_to_jtree(moralize(bnet15.dag), ones(1,N), stages, clusters);
wolffd@0: 
wolffd@0: clq_len = [];
wolffd@0: for c=1:length(cliques)
wolffd@0:   clq_len(c) = length(cliques{c});
wolffd@0: end
wolffd@0: hist(clq_len, 1:max(clq_len));
wolffd@0: h=hist(clq_len, 1:max(clq_len));
wolffd@0: axis([1 max(clq_len)+1 0 max(h)+1])
wolffd@0: xlabel('clique size','fontsize',16)
wolffd@0: ylabel('number','fontsize',16)
wolffd@0: 
wolffd@0: 
wolffd@0: 
wolffd@0: