Bayesian Networks in the Absence of Temporal Information - Bayesian Networks in R: With Applications in Systems Biology - page 43

Grow−Shrink Hill−climbing

ANL ANL

STAT STAT

ALG ALG

VECT VECT

MECH MECH

Fig. 2.6 Bayesian networks learned from the discretized marks data using Grow-Shrink (on the
left ) and hill-climbing (on the right ). Both belong to the same equivalence class and thus share the
same CPDAG representation

> fitted = bn.fit(bn.dhc, data = dmarks)
> fitted$ALG

Parameters of node ALG (multinomial distribution)

Conditional probability table:

VECT
ALG [8.93,45.5] (45.5,82.1]
[14.9,47.5] 0.5806452 0.2280702
(47.5,80.1] 0.4193548 0.7719298
Now that all the variables are discrete, we can also use package catnet ,which
does not handle continuous data. Since we do not know the correct topologi-
cal ordering of the nodes, we will need to call both the cnSearchSA and the
cnFindBIC functions. The former performs a simulated annealing search in the
space of orderings, returning the networks with the highest likelihood given their
respective node orderings; the latter then returns the network with the highest BIC
score among them.

> netlist = cnSearchSA(dmarks)
> best = cnFindBIC(netlist, nrow(dmarks))
> cnMatEdges(best)
[,1] [,2]
[1,] "MECH" "VECT"
[2,] "VECT" "ALG"
[3,] "ALG" "ANL"
[4,] "ALG" "STAT"