Mercurial > hg > camir-aes2014

annotate toolboxes/FullBNT-1.0.7/bnt/examples/dynamic/HHMM/Map/Old/mk_map_hhmm.m @ 0:e9a9cd732c1e tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
rev   line source
wolffd@0 1 function bnet = mk_map_hhmm(varargin)
wolffd@0 2
wolffd@0 3 % p is the prob of a successful move (defines the reliability of motors)
wolffd@0 4 p = 1;
wolffd@0 5 num_obs_nodes = 1;
wolffd@0 6
wolffd@0 7 for i=1:2:length(varargin)
wolffd@0 8 switch varargin{i},
wolffd@0 9 case 'p', p = varargin{i+1};
wolffd@0 10 case 'numobs', num_obs_node = varargin{i+1};
wolffd@0 11 end
wolffd@0 12 end
wolffd@0 13
wolffd@0 14
wolffd@0 15 q = 1-p;
wolffd@0 16
wolffd@0 17 % assign numbers to the nodes in topological order
wolffd@0 18 U = 1; A = 2; C = 3; F = 4; O = 5;
wolffd@0 19
wolffd@0 20 % create graph structure
wolffd@0 21
wolffd@0 22 ss = 5; % slice size
wolffd@0 23 intra = zeros(ss,ss);
wolffd@0 24 intra(U,F)=1;
wolffd@0 25 intra(A,[C F O])=1;
wolffd@0 26 intra(C,[F O])=1;
wolffd@0 27
wolffd@0 28 inter = zeros(ss,ss);
wolffd@0 29 inter(U,[A C])=1;
wolffd@0 30 inter(A,[A C])=1;
wolffd@0 31 inter(F,[A C])=1;
wolffd@0 32 inter(C,C)=1;
wolffd@0 33
wolffd@0 34 % node sizes
wolffd@0 35 ns = zeros(1,ss);
wolffd@0 36 ns(U) = 2; % left/right
wolffd@0 37 ns(A) = 2;
wolffd@0 38 ns(C) = 3;
wolffd@0 39 ns(F) = 2;
wolffd@0 40 ns(O) = 5; % we will assign each state a unique symbol
wolffd@0 41 l = 1; r = 2; % left/right
wolffd@0 42 L = 1; R = 2;
wolffd@0 43
wolffd@0 44 % Make the DBN
wolffd@0 45 bnet = mk_dbn(intra, inter, ns, 'observed', O);
wolffd@0 46 eclass = bnet.equiv_class;
wolffd@0 47
wolffd@0 48
wolffd@0 49
wolffd@0 50 % Define CPDs for slice 1
wolffd@0 51 % We clamp all of them, i.e., do not try to learn them.
wolffd@0 52
wolffd@0 53 % uniform probs over actions (the input could be chosen from a policy)
wolffd@0 54 bnet.CPD{eclass(U,1)} = tabular_CPD(bnet, U, 'CPT', mk_stochastic(ones(ns(U),1)), ...
wolffd@0 55 'adjustable', 0);
wolffd@0 56
wolffd@0 57 % uniform probs over starting abstract state
wolffd@0 58 bnet.CPD{eclass(A,1)} = tabular_CPD(bnet, A, 'CPT', mk_stochastic(ones(ns(A),1)), ...
wolffd@0 59 'adjustable', 0);
wolffd@0 60
wolffd@0 61 % Uniform probs over starting concrete state, modulo the fact
wolffd@0 62 % that corridor 2 is only of length 2.
wolffd@0 63 CPT = zeros(ns(A), ns(C)); % CPT(i,j) = P(C starts in j | A=i)
wolffd@0 64 CPT(1, :) = [1/3 1/3 1/3];
wolffd@0 65 CPT(2, :) = [1/2 1/2 0];
wolffd@0 66 bnet.CPD{eclass(C,1)} = tabular_CPD(bnet, C, 'CPT', CPT, 'adjustable', 0);
wolffd@0 67
wolffd@0 68 % Termination probs
wolffd@0 69 CPT = zeros(ns(U), ns(A), ns(C), ns(F));
wolffd@0 70 CPT(r,1,1,:) = [1 0];
wolffd@0 71 CPT(r,1,2,:) = [1 0];
wolffd@0 72 CPT(r,1,3,:) = [q p];
wolffd@0 73 CPT(r,2,1,:) = [1 0];
wolffd@0 74 CPT(r,2,2,:) = [q p];
wolffd@0 75 CPT(l,1,1,:) = [q p];
wolffd@0 76 CPT(l,1,2,:) = [1 0];
wolffd@0 77 CPT(l,1,3,:) = [1 0];
wolffd@0 78 CPT(l,2,1,:) = [q p];
wolffd@0 79 CPT(l,2,2,:) = [1 0];
wolffd@0 80
wolffd@0 81 bnet.CPD{eclass(F,1)} = tabular_CPD(bnet, F, 'CPT', CPT);
wolffd@0 82
wolffd@0 83
wolffd@0 84 % Assign each state a unique observation
wolffd@0 85 CPT = zeros(ns(A), ns(C), ns(O));
wolffd@0 86 CPT(1,1,1)=1;
wolffd@0 87 CPT(1,2,2)=1;
wolffd@0 88 CPT(1,3,3)=1;
wolffd@0 89 CPT(2,1,4)=1;
wolffd@0 90 CPT(2,2,5)=1;
wolffd@0 91 %CPT(2,3,:) undefined
wolffd@0 92
wolffd@0 93 bnet.CPD{eclass(O,1)} = tabular_CPD(bnet, O, 'CPT', CPT);
wolffd@0 94
wolffd@0 95
wolffd@0 96 % Define the CPDs for slice 2
wolffd@0 97
wolffd@0 98 % Abstract
wolffd@0 99
wolffd@0 100 % Since the top level never resets, the starting distribution is irrelevant:
wolffd@0 101 % A2 will be determined by sampling from transmat(A1,:).
wolffd@0 102 % But the code requires we specify it anyway; we make it all 0s, a dummy value.
wolffd@0 103 startprob = zeros(ns(U), ns(A));
wolffd@0 104
wolffd@0 105 transmat = zeros(ns(U), ns(A), ns(A));
wolffd@0 106 transmat(R,1,:) = [q p];
wolffd@0 107 transmat(R,2,:) = [0 1];
wolffd@0 108 transmat(L,1,:) = [1 0];
wolffd@0 109 transmat(L,2,:) = [p q];
wolffd@0 110
wolffd@0 111 % Qps are the parents we condition the parameters on, in this case just
wolffd@0 112 % the past action.
wolffd@0 113 bnet.CPD{eclass(A,2)} = hhmm2Q_CPD(bnet, A+ss, 'Fbelow', F, ...
wolffd@0 114 'startprob', startprob, 'transprob', transmat);
wolffd@0 115
wolffd@0 116
wolffd@0 117
wolffd@0 118 % Concrete
wolffd@0 119
wolffd@0 120 transmat = zeros(ns(C), ns(U), ns(A), ns(C));
wolffd@0 121 transmat(1,r,1,:) = [q p 0.0];
wolffd@0 122 transmat(2,r,1,:) = [0.0 q p];
wolffd@0 123 transmat(3,r,1,:) = [0.0 0.0 1.0];
wolffd@0 124 transmat(1,r,2,:) = [q p 0.0];
wolffd@0 125 transmat(2,r,2,:) = [0.0 1.0 0.0];
wolffd@0 126 %
wolffd@0 127 transmat(1,l,1,:) = [1.0 0.0 0.0];
wolffd@0 128 transmat(2,l,1,:) = [p q 0.0];
wolffd@0 129 transmat(3,l,1,:) = [0.0 p q];
wolffd@0 130 transmat(1,l,2,:) = [1.0 0.0 0.0];
wolffd@0 131 transmat(2,l,2,:) = [p q 0.0];
wolffd@0 132
wolffd@0 133 % Add a new dimension for A(t-1), by copying old vals,
wolffd@0 134 % so the matrix is the same size as startprob
wolffd@0 135
wolffd@0 136
wolffd@0 137 transmat = reshape(transmat, [ns(C) ns(U) ns(A) 1 ns(C)]);
wolffd@0 138 transmat = repmat(transmat, [1 1 1 ns(A) 1]);
wolffd@0 139
wolffd@0 140 % startprob(C(t-1), U(t-1), A(t-1), A(t), C(t))
wolffd@0 141 startprob = zeros(ns(C), ns(U), ns(A), ns(A), ns(C));
wolffd@0 142 startprob(1,L,1,1,:) = [1.0 0.0 0.0];
wolffd@0 143 startprob(3,R,1,2,:) = [1.0 0.0 0.0];
wolffd@0 144 startprob(3,R,1,1,:) = [0.0 0.0 1.0];
wolffd@0 145 %
wolffd@0 146 startprob(1,L,2,1,:) = [0.0 0.0 010];
wolffd@0 147 startprob(2,L,2,1,:) = [1.0 0.0 0.0];
wolffd@0 148 startprob(2,R,2,2,:) = [0.0 1.0 0.0];
wolffd@0 149
wolffd@0 150 % want transmat(U,A,C,At,Ct), ie. in topo order
wolffd@0 151 transmat = permute(transmat, [2 3 1 4 5]);
wolffd@0 152 startprob = permute(startprob, [2 3 1 4 5]);
wolffd@0 153 bnet.CPD{eclass(C,2)} = hhmm2Q_CPD(bnet, C+ss, 'Fself', F, ...
wolffd@0 154 'startprob', startprob, 'transprob', transmat);
wolffd@0 155
wolffd@0 156