gRbase (version 1.7-5)

triangulate: Triangulation of an undirected graph

Description

This function will triangulate an undirected graph by adding fill-ins.

Usage

triangulate(object, ...)
"triangulate"(object, nLevels = NULL, result=NULL,
check=TRUE, ...)
triangulateMAT(amat, nLevels=rep(2, ncol(amat)), ...)

Arguments

object

An undirected graph represented either as a graphNEL object, an igraph, a (dense) matrix, a (sparse) dgCMatrix.

nLevels

The number of levels of the variables (nodes) when these are discrete. Used in determining the triangulation using a "minimum clique weight heuristic". See section 'details'.

result

The type (representation) of the result. Possible values are "graphNEL", "igraph", "matrix", "dgCMatrix". Default is the same as the type of object.

check

If TRUE (the default) it is checked whether the graph is triangulated before doing the triangulation; gives a speed up

...

Additional arguments, currently not used.

amat

Adjacency matrix; a (dense) matrix, or a (sparse) dgCMatrix.

Value

graphNEL, a (dense) matrix or a (sparse) dgCMatrix.

Details

The workhorse is the triangulateMAT function.

The triangulation is made so as the total state space is kept low by applying a minimum clique weight heuristic: When a fill-in is necessary, the algorithm will search for an edge to add such that the complete set to be formed will have as small a state-space as possible. It is in this connection that the nLevels values are used.

Default (when nLevels=NULL) is to take nLevels=2 for all nodes. If nLevels is the same for all nodes then the heuristic aims at keeping the clique sizes small.

See Also

ug dag mcs, mcsMAT rip, ripMAT, moralize, moralizeMAT

Examples

## graphNEL
uG1 <- ug(~a:b+b:c+c:d+d:e+e:f+f:a)
tuG1 <- triangulate(uG1)

## adjacency matrix
uG2 <- ug(~a:b+b:c+c:d+d:e+e:f+f:a, result="matrix")
tuG2 <- triangulate(uG2)

## adjacency matrix (sparse)
uG2 <- ug(~a:b+b:c+c:d+d:e+e:f+f:a, result="Matrix")
tuG2 <- triangulate(uG2)