Daniel@0: function bnet = mk_chmm(N, Q, Y, discrete_obs, coupled, CPD)
Daniel@0: % MK_CHMM Make a coupled Hidden Markov Model
Daniel@0: %
Daniel@0: % There are N hidden nodes, each connected to itself and its two nearest neighbors in the next
Daniel@0: % slice (apart from the edges, where there is 1 nearest neighbor).
Daniel@0: %
Daniel@0: % Example: If N = 3, the hidden backbone is as follows, where all arrows point to the righ+t
Daniel@0: %
Daniel@0: % X1--X2
Daniel@0: %   \/ 
Daniel@0: %   /\
Daniel@0: % X2--X2
Daniel@0: %   \/ 
Daniel@0: %   /\
Daniel@0: % X3--X3
Daniel@0: %
Daniel@0: % Each hidden node has a "private" observed child (not shown).
Daniel@0: %
Daniel@0: % BNET = MK_CHMM(N, Q, Y)
Daniel@0: % Each hidden node is discrete and has Q values.
Daniel@0: % Each observed node is a Gaussian vector of length Y.
Daniel@0: %
Daniel@0: % BNET = MK_CHMM(N, Q, Y, DISCRETE_OBS)
Daniel@0: % If discrete_obs = 1, the observations are discrete (values in {1, .., Y}).
Daniel@0: %
Daniel@0: % BNET = MK_CHMM(N, Q, Y, DISCRETE_OBS, COUPLED)
Daniel@0: % If coupled = 0, the chains are not coupled, i.e., we make N parallel HMMs.
Daniel@0: %
Daniel@0: % BNET = MK_CHMM(N, Q, Y, DISCRETE_OBS, COUPLED, CPDs)
Daniel@0: % means use the specified CPD structures instead of creating random params.
Daniel@0: %  CPD{i}.CPT, i=1:N specifies the prior
Daniel@0: %  CPD{i}.CPT, i=2N+1:3N specifies the transition model
Daniel@0: %  CPD{i}.mean, CPD{i}.cov, i=N+1:2N specifies the observation model if Gaussian
Daniel@0: %  CPD{i}.CPT, i=N+1:2N if discrete
Daniel@0: 
Daniel@0: 
Daniel@0: if nargin < 2, Q = 2; end
Daniel@0: if nargin < 3, Y = 1; end
Daniel@0: if nargin < 4, discrete_obs = 0; end
Daniel@0: if nargin < 5, coupled = 1; end
Daniel@0: if nargin < 6, rnd = 1; else rnd = 0; end
Daniel@0:   
Daniel@0: ss = N*2;
Daniel@0: hnodes = 1:N;
Daniel@0: onodes = (1:N)+N;
Daniel@0: 
Daniel@0: intra = zeros(ss);
Daniel@0: for i=1:N
Daniel@0:   intra(hnodes(i), onodes(i))=1;
Daniel@0: end
Daniel@0: 
Daniel@0: inter = zeros(ss);
Daniel@0: if coupled
Daniel@0:   for i=1:N
Daniel@0:     inter(i, max(i-1,1):min(i+1,N))=1;
Daniel@0:   end
Daniel@0: else
Daniel@0:   inter(1:N, 1:N) = eye(N);
Daniel@0: end  
Daniel@0: 
Daniel@0: ns = [Q*ones(1,N) Y*ones(1,N)]; 
Daniel@0: 
Daniel@0: eclass1 = [hnodes onodes];
Daniel@0: eclass2 = [hnodes+ss onodes];
Daniel@0: if discrete_obs
Daniel@0:   dnodes = 1:ss;
Daniel@0: else
Daniel@0:   dnodes = hnodes;
Daniel@0: end
Daniel@0: bnet = mk_dbn(intra, inter, ns, 'discrete', dnodes, 'eclass1', eclass1, 'eclass2', eclass2, ...
Daniel@0: 	      'observed', onodes);
Daniel@0: 
Daniel@0: if rnd
Daniel@0:   for i=hnodes(:)'
Daniel@0:     bnet.CPD{i} = tabular_CPD(bnet, i);
Daniel@0:   end
Daniel@0:   for i=onodes(:)'
Daniel@0:     if discrete_obs
Daniel@0:       bnet.CPD{i} = tabular_CPD(bnet, i);
Daniel@0:     else
Daniel@0:       bnet.CPD{i} = gaussian_CPD(bnet, i);
Daniel@0:     end
Daniel@0:   end
Daniel@0:   for i=hnodes(:)'+ss
Daniel@0:     bnet.CPD{i} = tabular_CPD(bnet, i);
Daniel@0:   end
Daniel@0: else
Daniel@0:   for i=hnodes(:)'
Daniel@0:     bnet.CPD{i} = tabular_CPD(bnet, i, CPD{i}.CPT);
Daniel@0:   end
Daniel@0:   for i=onodes(:)'
Daniel@0:     if discrete_obs
Daniel@0:       bnet.CPD{i} = tabular_CPD(bnet, i, CPD{i}.CPT);
Daniel@0:     else
Daniel@0:       bnet.CPD{i} = gaussian_CPD(bnet, i, CPD{i}.mean, CPD{i}.cov);
Daniel@0:     end
Daniel@0:   end
Daniel@0:   for i=hnodes(:)'+ss
Daniel@0:     bnet.CPD{i} = tabular_CPD(bnet, i, CPD{i}.CPT);
Daniel@0:   end
Daniel@0: end
Daniel@0: 
Daniel@0: