Font Size:
A GROUP LASSO APPROACH FOR SELECTING THE NUMBER OF STATES IN A HIDDEN MARKOV MODEL
Last modified: 2023-07-08
Abstract
In hidden Markov models (HMMs), the selection of an adequate number of states is commonly based on information criteria, despite well-known problems and pitfalls. We explore an alternative approach, considering a penalised likelihood
comprising a group lasso penalty on the entries of the transition probability matrix. The feasibility of the approach is assessed in two simulation experiments, where we investigate how often the suggested approach accurately selects the true number of states compared to the common benchmarks.
comprising a group lasso penalty on the entries of the transition probability matrix. The feasibility of the approach is assessed in two simulation experiments, where we investigate how often the suggested approach accurately selects the true number of states compared to the common benchmarks.