wolffd@0: function [int, persist, transient] = compute_interface_nodes(intra, inter)
wolffd@0: % COMPUTE_INTERFACE_NODES Find the nodes in a DBN that represent a sufficient statistic
wolffd@0: % [int, persist, transient] = compute_interface_nodes(intra, inter)
wolffd@0: %
wolffd@0: % The interface nodes are all those that has an incoming temporal arc,
wolffd@0: % or which have a child which has an incoming temporal arc,
wolffd@0: % where a temporal arc means one coming from the previous slice.
wolffd@0: % (The parents of nodes with incoming temporal arcs are needed
wolffd@0: % because moralization will bring them into the clique.)
wolffd@0: %
wolffd@0: % The persisent nodes are all those that have one or more incoming temporal arc.
wolffd@0: % The transient nodes are all the non-persistent.
wolffd@0: %
wolffd@0: % See U. Kjaerulff, "dHugin: A computational system for dynamic
wolffd@0: % time-sliced Bayesian networks", Intl. J. Forecasting (11) 89-111, 1995
wolffd@0: 
wolffd@0: n = length(intra);
wolffd@0: int = [];
wolffd@0: persist = [];
wolffd@0: for u=1:n
wolffd@0:   if any(inter(:,u))
wolffd@0:     int = [int u];
wolffd@0:     persist = [persist u];
wolffd@0:   end
wolffd@0:   if any(inter(:, children(intra, u)))
wolffd@0:     int = [int u];
wolffd@0:   end
wolffd@0: end
wolffd@0: int = unique(int);
wolffd@0: persist = unique(persist);
wolffd@0: transient = mysetdiff(1:n, persist);