Mercurial > hg > camir-aes2014

comparison toolboxes/FullBNT-1.0.7/bnt/examples/dynamic/ho1.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
comparison
equal deleted inserted replaced
-1:000000000000 0:e9a9cd732c1e
1 function ho1()
2
3 % Example of how to create a higher order DBN
4 % Written by Rainer Deventer <deventer@informatik.uni-erlangen.de> 3/28/03
5
6 bnet = createBNetNL();
7
8 %%%%%%%%%%%%
9
10
11 function bnet = createBNetNL(varargin)
12 % Generate a Bayesian network, which is able to model nonlinearities at
13 % the input. The only input is the order of the dynamic system. If this
14 % parameter is missing, the an order of two is assumed
15 if nargin > 0
16 order = varargin{1}
17 else
18 order = 2;
19 end
20
21 ss = 6; % For each time slice the following nodes are modeled
22 % ud(t_k) Discrete node, which decides whether saturation is reached.
23 % Node number 2
24 % uv(t_k) Visible input node with node number 2
25 % uh(t_k) Hidden input node with node number 3
26 % y(t_k) Modeled output, Number 4
27 % z(t_k) Disturbing variable, number 5
28 % q(t_k), number6 6
29
30 intra = zeros(ss,ss);
31 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
32 % Within each timeslice ud(t_k) is connected with uv(t_k) and uh(t_k) %
33 % This part is used to model saturation %
34 % A connection from uv(t_k) to uh(t_k) is omitted %
35 % Additionally y(t_k) is connected with q(t_k). To model the disturbing%
36 % value z(t_k) is connected with q(t_k). %
37 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
38 intra(1,2:3) = 1; % Connections ud(t_k) -> uv(t_k) and ud(t_k) -> uh(t_k)
39 intra(4:5,6) = 1; % Connectios y(t_k) -> q(t_k) and z(t_k) -> q(t_k)
40
41
42
43 inter = zeros(ss,ss,order);
44 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
45 % The Markov assumption is not met as connections from time slice t to t+2 %
46 % exist. %
47 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
48 for i = 1:order
49 if i == 1
50 inter(1,1,i) = 1; %Connect the discrete nodes. This is necessary to improve
51 %the disturbing reaction
52 inter(3,4,i) = 1; %Connect uh(t_{k-1}) with y(t_k)
53 inter(4,4,i) = 1; %Connect y(t_{k-1}) with y(t_k)
54 inter(5,5,i) = 1; %Connect z(t_{k-1}) with z(t_k)
55 else
56 inter(3,4,i) = 1; %Connect uh(t_{k-i}) with y(t_k)
57 inter(4,4,i) = 1; %Connect y(t_{k-i}) with y(t_k)
58 end
59 end
60
61 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
62 % Define the dimensions of the discrete nodes. Node 1 has two states %
63 % 1 = lower saturation reached %
64 % 2 = Upper saturation reached %
65 % Values in between are model by probabilities between 0 and 1 %
66 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
67 node_sizes = ones(1,ss);
68 node_sizes(1) = 2;
69 dnodes = [1];
70
71 eclass = [1:6;7 2:3 8 9 6;7 2:3 10 11 6];
72 bnet = mk_higher_order_dbn(intra,inter,node_sizes,...
73 'discrete',dnodes,...
74 'eclass',eclass);
75
76 cov_high = 400;
77 cov_low = 0.01;
78 weight1 = randn(1,1);
79 weight2 = randn(1,1);
80 weight3 = randn(1,1);
81 weight4 = randn(1,1);
82
83 numOfNodes = 5 + order;
84 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
85 % Nodes of the first time-slice %
86 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
87 % Discrete input node,
88 bnet.CPD{1} = tabular_CPD(bnet,1,'CPT',[1/2 1/2],'adjustable',0);
89
90
91 % Modeled visible input
92 bnet.CPD{2} = gaussian_CPD(bnet,2,'mean',[0 10],'clamp_mean',1,...
93 'cov',[10 10],'clamp_cov',1);
94
95 % Modeled hidden input
96 bnet.CPD{3} = gaussian_CPD(bnet,3,'mean',[0, 10],'clamp_mean',1,...
97 'cov',[0.1 0.1],'clamp_cov',1);
98
99 % Modeled output in the first timeslice, thus there are no parents
100 % Usuallz the output nodes get a low covariance. But in the first
101 % time-slice a prediction of the output is not possible due to
102 % missing information
103 bnet.CPD{4} = gaussian_CPD(bnet,4,'mean',0,'clamp_mean',1,...
104 'cov',cov_high,'clamp_cov',1);
105
106 %Disturbance
107 bnet.CPD{5} = gaussian_CPD(bnet,5,'mean',0,...
108 'cov',[4],...
109 'clamp_mean',1,...
110 'clamp_cov',1);
111
112 %Observed output.
113 bnet.CPD{6} = gaussian_CPD(bnet,6,'mean',0,...
114 'clamp_mean',1,...
115 'cov',cov_low,'clamp_cov',1,...
116 'weights',[1 1],'clamp_weights',1);
117
118 % Discrete node at second time slice
119 bnet.CPD{7} = tabular_CPD(bnet,7,'CPT',[0.6 0.4 0.4 0.6],'adjustable',0);
120 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
121 % Node for the model output %
122 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
123 bnet.CPD{8} = gaussian_CPD(bnet,10,'mean',0,...
124 'cov',cov_high,...
125 'clamp_mean',1,...
126 'clamp_cov',1);
127 % 'weights',[0.0791 0.9578]);
128
129
130 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
131 % Node for the disturbance %
132 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
133 bnet.CPD{9} = gaussian_CPD(bnet,11,'mean',0,'clamp_mean',1,...
134 'cov',[4],'clamp_cov',1,...
135 'weights',[1],'clamp_weights',1);
136
137 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
138 % Node for the model output %
139 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
140 bnet.CPD{10} = gaussian_CPD(bnet,16,'mean',0,'clamp_mean',1,...
141 'cov',cov_low,'clamp_cov',1);
142 % 'weights',[0.0188 -0.0067 0.0791 0.9578]);
143
144
145
146
147 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
148 % Node for the disturbance %
149 %%%%%%%%%%%%%%%%%%%%%%%%%%%%
150 bnet.CPD{11} = gaussian_CPD(bnet,17,'mean',0,'clamp_mean',1,...
151 'cov',[0.2],'clamp_cov',1,...
152 'weights',[1],'clamp_weights',1);
153
154
155
156