mhmm: Mixture model of Hidden Markov Models.

In MHMM: Finite Mixture of Hidden Markov Model

Description

This function performs maximum likelihood inference of mixture of hidden Markov models

Usage

mhmm(y, K, M, smartinit = TRUE, nbinit = 100, tol = 10^(-4),
  nbKeep = min(20, nbinit), iterSmall = 10, nbcores = 1)

Arguments

y	list. Observations (one element of the list per subject, missing values should be encoded by NA).
K	numeric. Number of mixture components.
M	numeric. Number of states.
smartinit	boolean. If TRUE, MLE of the mixture neglecting the time dependency is used for the initialization of the algorithm.
nbinit	numeric.Number of initializations.
tol	numeric. It indicates the maximal gap between two successive iterations of EM algorithm which stops the algorithm
nbKeep	numeric. It indicates the number of chains used for the final EM algorithm
iterSmall	numeric. It indicates the number of iterations for each SmallEM algorithm
nbcores	numeric. It defines the numerber of cores used by the algorithme

Value

Returns an instance of mhmmresults.

References

Du Roy de Chaumaray, M. and Marbac, M. and Navarro, F. (2019). Mixture of hidden Markov models for accelerometer data. arXiv preprint arXiv:1906.01547

Examples

data(accelero)
# To make the estimation <5
res <- mhmm(accelero, K = 2, M = 4, nbcores = 1, nbinit = 5, iterSmall = 2)
plot(res, 1)

 
data(accelero)
# It is better to increase the number of random initializations
res <- mhmm(accelero, K = 2, M = 4, nbcores = 1)
plot(res, 1)

MHMM documentation built on March 26, 2020, 7:46 p.m.