baumWelchRecursion: Implementation of the Baum Welch Algorithm as a special case...

In dagHMM: Directed Acyclic Graph HMM with TAN Structured Emissions

Implementation of the Baum Welch Algorithm as a special case of EM algorithm

Description

baumWelch recursively calls this function to give a final estimate of parameters for dag HMM Uses Parallel Processing to speed up calculations for large data. Should not be used directly.

Usage

baumWelchRecursion(hmm, observation, t_seq, kn_states = NULL, kn_verify = NULL)

Arguments

hmm	hmm Object of class List given as output by initHMM
observation	Dataframe containing the discritized character values of only covariates at each node. Column names of dataframe should be same as the covariate names. Missing values should be denoted by "NA".
t_seq	list of forward and backward DAG traversal sequence (in that order) as given output by fwd_seq_gen and bwd_seq_gen function
kn_states	(Optional) A (L * 2) dataframe where L is the number of training nodes where state values are known. First column should be the node number and the second column being the corresponding known state values of the nodes
kn_verify	(Optional) A (L * 2) dataframe where L is the number of validation nodes where state values are known. First column should be the node number and the second column being the corresponding known state values of the nodes

Value

List containing estimated Transition and Emission probability matrices

See Also

baumWelch

Examples

library(bnlearn)

tmat = matrix(c(0,0,1,0,0,0,0,1,0,0,0,0,0,1,1,0,0,0,0,0,0,0,0,0,0),
               5,5, byrow= TRUE ) #for "X" (5 nodes) shaped dag
states = c("P","N") #"P" represent cases(or positive) and "N" represent controls(or negative)
bnet = model2network("[A][C|A:B][D|A:C][B|A]") #A is the target variable while
                                               #B, C and D are covariates
obsvA=data.frame(list(B=c("L","H","H","L","L"),C=c("H","H","L","L","H"),D=c("L","L","L","H","H")))
hmmA = initHMM(States=states, dagmat= tmat, net=bnet, observation=obsvA)
kn_st = data.frame(node=c(2),state=c("P"),stringsAsFactors = FALSE)
                   #state at node 2 is known to be "P"
kn_vr = data.frame(node=c(3,4,5),state=c("P","N","P"),stringsAsFactors = FALSE)
                   #state at node 3,4,5 are "P","N","P" respectively
t_seq=list(fwd_seq_gen(hmmA),bwd_seq_gen(hmmA))
newparam= baumWelchRecursion(hmm=hmmA,observation=obsvA,t_seq=t_seq,
                             kn_states=kn_st, kn_verify=kn_vr)

dagHMM documentation built on Jan. 11, 2023, 1:13 a.m.