plotHMM (version 2023.8.28)

forward_interface: Forward algorithm

Description

Efficient implementation of forward algorithm in C++ code, for N data and S states.

Usage

forward_interface(
  log_emission_mat, log_transition_mat, log_initial_prob_vec)

Value

list with two elements

log_alpha

N x S numeric matrix of forward log likelihood at each data/state.

log_lik

numeric scalar total log likelihood of data given model parameters.

Arguments

log_emission_mat

N x S numeric matrix of log likelihood of observing each data point in each state.

log_transition_mat

S x S numeric matrix; log_transition_mat[i,j] is the log probability of going from state i to state j.

log_initial_prob_vec

S numeric vector of log probabilities of observing each state at the beginning of the sequence.

Author

Toby Dylan Hocking

Examples

##simulated data.
seg.mean.vec <- c(2, 0, -1, 0)
data.mean.vec <- rep(seg.mean.vec, each=10)
set.seed(1)
N.data <- length(data.mean.vec)
y.vec <- rnorm(N.data, data.mean.vec)
##model.
n.states <- 3
log.A.mat <- log(matrix(1/n.states, n.states, n.states))
state.mean.vec <- c(-1, 0, 1)*0.1
sd.param <- 1
log.pi.vec <- log(rep(1/n.states, n.states))
log.emission.mat <- dnorm(
  y.vec,
  matrix(state.mean.vec, N.data, n.states, byrow=TRUE),
  sd.param,
  log=TRUE)
plotHMM::forward_interface(log.emission.mat, log.A.mat, log.pi.vec)