Interfacing with the igraph R package

The igraph package (link) is the R interface to the igraph library (link) for network analysis. It implements an extensive selection of algorithms for creating and generating directed and undirected graphs, manipulating nodes and arcs, and it provides highly customizable plotting facilities. For this reason, bnlearn implements functions to import and export network structures to igraph's native objects, which are themselves of class igraph.

Exporting a network structure to igraph

Exporting objects is achieved with the conversion method as.igraph().

> library(bnlearn)
> 
> dag.bnlearn = random.graph(LETTERS[1:10])
> dag.bnlearn


  Random/Generated Bayesian network

  model:
   [A][B][C][F][G][D|C][H|C:F][E|D][I|A:F:H][J|E] 
  nodes:                                 10 
  arcs:                                  8 
    undirected arcs:                     0 
    directed arcs:                       8 
  average markov blanket size:           2.20 
  average neighbourhood size:            1.60 
  average branching factor:              0.80 

  generation algorithm:                  Full Ordering 
  arc sampling probability:              0.2222222

> dag.igraph = as.igraph(dag.bnlearn)
> dag.igraph

IGRAPH fc6706b DN-- 10 8 -- 
+ attr: name (v/c)
+ edges from fc6706b (vertex names):
[1] A->I C->D C->H D->E E->J F->H F->I H->I

as.igraph() accepts objects of class bn.fit as well as of class bn, and it implitly call bn.net() (link) on the former.

> dag.test = hc(learning.test)
> dag.test


  Bayesian network learned via Score-based methods

  model:
   [A][C][F][B|A][D|A:C][E|B:F] 
  nodes:                                 6 
  arcs:                                  5 
    undirected arcs:                     0 
    directed arcs:                       5 
  average markov blanket size:           2.33 
  average neighbourhood size:            1.67 
  average branching factor:              0.83 

  learning algorithm:                    Hill-Climbing 
  score:                                 BIC (disc.) 
  penalization coefficient:              4.258597 
  tests used in the learning procedure:  40 
  optimized:                             TRUE

> fitted = bn.fit(dag.test, learning.test)
> as.igraph(fitted)

IGRAPH 25a1eb4 DN-- 6 5 -- 
+ attr: name (v/c)
+ edges from 25a1eb4 (vertex names):
[1] A->B A->D B->E C->D F->E

Undirected arcs are exported as a pair of directed arcs, that is, Xi — Xj becomes {Xi → Xj, Xi ← Xj}.

> as.igraph(skeleton(dag.test))

IGRAPH aa58703 DN-- 6 10 -- 
+ attr: name (v/c)
+ edges from aa58703 (vertex names):
 [1] A->B A->D B->A B->E C->D D->A D->C E->B E->F F->E

Importing a network structure from igraph

Importing network structures works in the same way, with a conversion method as.bn().

> as.bn(dag.igraph)


  Random/Generated Bayesian network

  model:
   [A][B][C][F][G][D|C][H|C:F][E|D][I|A:F:H][J|E] 
  nodes:                                 10 
  arcs:                                  8 
    undirected arcs:                     0 
    directed arcs:                       8 
  average markov blanket size:           2.20 
  average neighbourhood size:            1.60 
  average branching factor:              0.80 

  generation algorithm:                  Empty

Note that as graphs are imported with as.bn() they are checked to be acyclic, unless the user specifies check.cycles = FALSE.

> library(igraph)
> 
> cyclic.graph = make_graph(~ A-+B-+C-+D-+A)
> cyclic.graph

IGRAPH 8929438 DN-- 4 4 -- 
+ attr: name (v/c)
+ edges from 8929438 (vertex names):
[1] A->B B->C C->D D->A

> as.bn(cyclic.graph)

## Error: the igraph object contains directed cycles.

> as.bn(cyclic.graph, check.cycles = FALSE)


  Random/Generated Bayesian network

  model:
   [A|D][B|A][C|B][D|C] 
  nodes:                                 4 
  arcs:                                  4 
    undirected arcs:                     0 
    directed arcs:                       4 
  average markov blanket size:           2.00 
  average neighbourhood size:            2.00 
  average branching factor:              1.00 

  generation algorithm:                  Empty
Last updated on Mon Aug 5 02:49:23 2024 with bnlearn 5.0 and R version 4.4.1 (2024-06-14).