PBNT.Graph.JoinTree Class Reference

Inheritance diagram for PBNT.Graph.JoinTree:

Collaboration diagram for PBNT.Graph.JoinTree:

Public Member Functions
def	__init__ (self, clique)
def	init_clique_potentials (self, variables)
def	merge (self, sepset, tree)
def	reinitialize (self, variables)
def	enter_evidence (self, evidence, nodes)
Public Member Functions inherited from PBNT.Graph.Graph
def	__init__ (self, nodes)
def	add_node (self, node)
def	member_of (self, node)
def	contains (self, nodes)
def	connect_nodes (self, node1, node2)

Public Attributes
	initialized
	likelihoods
Public Attributes inherited from PBNT.Graph.Graph
	nodes

Detailed Description

JoinTree is the final graph that is constructed for JunctionTree Inference.  To create the JoinTree, we first create a forest of n JoinTrees where each tree consists of a single clique (n is the number of cliques).  Then we create a list of all distinct pairs.  Then we insert a sepset between each pair of cliques.  Then we loop n - 1 times.  At each iteration, we choose the next best sepset according to some heuristic.  If we the two cliques connected to the sepset are on different trees, we join them into one larger tree.

Member Function Documentation

def PBNT.Graph.JoinTree.enter_evidence	(		self,
			evidence,
			nodes
	)

For all nodes that are not blank, we want to make its family clique consistent with the evidence.  This can be done by setting all values of the clique that are consistent with the evidence to 1 and all other places to 0.

---

The documentation for this class was generated from the following file:

• PBNT/Graph.py