jtree_ndx_dbn_inf_engine.m

来自「麻省理工学院的人工智能工具箱,很珍贵,希望对大家有用!」· M 代码 · 共 131 行

function engine = jtree_ndx_dbn_inf_engine(bnet, varargin)
% JTREE_NDX_DBN_INF_ENGINE Junction tree inference algorithm for DBNs which pre-computes indices.
% engine = jtree_ndx_dbn_inf_engine(bnet, ...)
%
% This is to jtree_dbn_inf_engine what jtree_ndx_inf_engine is to jtree_inf_engine.
%
% Optional arguments
% ndx_type - 'B', 'D', or 'SD' [ 'SD' ]

ss = length(bnet.intra);

clusters = {};
ndx_type = 'SD';

if nargin >= 2
  args = varargin;
  nargs = length(args);
  for i=1:2:nargs
    switch args{i},
     case 'clusters', clusters = args{i+1};
     case 'ndx_type', ndx_type = args{i+1};
    end
  end
end

engine.evidence = [];
engine.node_sizes = [];
engine.ndx_type = ndx_type;

[int, engine.persist, engine.transient] = compute_interface_nodes(bnet.intra, bnet.inter);
%engine.interface = engine.persist; % WRONG!
engine.interface = int;
engine.nonint = mysetdiff(1:ss, int);

engine.bnet_parents = {};
for i=1:2*ss
  engine.bnet_parents{i} = parents(bnet.dag, i);
end

% figure out when we can speedup computation of the observation likelihood
engine.parents_in_same_slice = ones(1,ss);
engine.parents_hidden = ones(1,ss);
engine.time_invariant = ones(1,ss);
engine.tabular_node = zeros(1,ss);
engine.gaussian_node = zeros(1,ss);
for i=1:ss
  ps = parents(bnet.dag, i+ss);
  if any(ps <= ss)
    engine.parents_in_same_slice(i) = 0;
  end
  if mysubset(ps, onodes)
    engine.parents_hidden(i) = 0;
  end
  if bnet.eclass1(i) ~= bnet.eclass2(i)
    engine.time_invariant(i) = 0;
  end
  engine.simple_obs(i) = engine.parents_in_same_slice(i) & engine.parents_hidden(i) & ...
      engine.time_invariant(i);
  % the following is not always correct...
  engine.tabular_node(i) = myismember(i, bnet.dnodes); 
  engine.gaussian_node(i) = myismember(i, bnet.cnodes);
end
  
% Create a "1.5 slice" jtree, containing slice 1 and the interface nodes of slice 2
% To keep the node numbering the same, we simply disconnect the non-interface nodes
% from slice 2, and set their size to 1.
% We do this to speed things up, and so that the likelihood is computed correctly - we do not need to do
% this if we just want to compute marginals. 
intra15 = bnet.intra;
for i=engine.nonint(:)'
  intra15(i,:) = 0;
  intra15(:,i) = 0;
end
dag15 = [bnet.intra bnet.inter;
	 zeros(ss)    intra15];
ns = bnet.node_sizes(:);
ns(engine.nonint+ss) = 1; % disconnected nodes get size 1
obs_nodes = [onodes(:) onodes(:)+ss];
bnet15 = mk_bnet(dag15, ns, 'discrete', bnet.dnodes, 'equiv_class', bnet.equiv_class(:), ...
		 'observed', obs_nodes(:));
ns(obs_nodes) = 1;

% use unconstrained elimination,
% but force there to be a clique containing both interfaces
clusters(end+1:end+2) = {int, int+ss};
engine.jtree_engine = jtree_ndxSD_inf_engine(bnet15, 'clusters', clusters, 'root', int+ss);


engine.in_clq = clq_containing_nodes(engine.jtree_engine, int);
engine.out_clq = clq_containing_nodes(engine.jtree_engine, int+ss);
engine.jtree_struct = struct(engine.jtree_engine); % violate object privacy

cliques = engine.jtree_struct.cliques;

engine.mult_int_onto_inclq_ndx_id = add_ndxSD(cliques{engine.in_clq}, engine.interface, ns);
engine.marg_root_onto_int2_ndx_id = add_ndxSD(cliques{engine.jtree_struct.root_clq}, engine.interface+ss, ns);

engine.marg_inclq_onto_int_ndx_id = add_ndxSD(cliques{engine.in_clq}, engine.interface, ns);
engine.marg_outclq_onto_int2_ndx_id = add_ndxSD(cliques{engine.out_clq}, engine.interface+ss, ns);
engine.mult_int2_onto_outclq_ndx_id = add_ndxSD(cliques{engine.out_clq}, engine.interface+ss, ns);
engine.mult_int_onto_int_ndx_id = add_ndxSD(engine.interface, engine.interface, ns);




% Also create an engine just for slice 1
bnet1 = mk_bnet(bnet.intra1, bnet.node_sizes_slice, bnet.dnodes, bnet.equiv_class(:,1));
for i=1:max(bnet1.equiv_class)
  bnet1.CPD{i} = bnet.CPD{i};
end

engine.jtree_engine1 = jtree_ndxSD_inf_engine(bnet1, onodes, 'clusters', {int},  'root', int);

engine.in_clq1 = clq_containing_nodes(engine.jtree_engine1, int);
engine.jtree_struct1 = struct(engine.jtree_engine1); % violate object privacy

cliques = engine.jtree_struct1.cliques;
engine.mult_int_onto_inclq1_ndx_id = add_ndxSD(cliques{engine.in_clq1}, engine.interface, ns);
engine.marg_inclq1_onto_int_ndx_id = add_ndxSD(cliques{engine.in_clq1}, engine.interface, ns);


% stuff needed by marginal_nodes
engine.clpot = [];
engine.T = [];
engine.maximize = [];
engine.actual_node_sizes = [];
engine.eff_node_sizes = [];

engine = class(engine, 'jtree_ndxSD_dbn_inf_engine', inf_engine(bnet));