Since there's already a Bayesian Mixture model implementation, it would be great if someone could extend the HMM implementation to include mixture emissions.
On a side note, is the n_hidden parameter in the HMM model deterministic or a maximum ala truncated HDP-HMM models? If the former, automatic state discovery would be nice but probably a lot more work.
Since there's already a Bayesian Mixture model implementation, it would be great if someone could extend the HMM implementation to include mixture emissions.
On a side note, is the n_hidden parameter in the HMM model deterministic or a maximum ala truncated HDP-HMM models? If the former, automatic state discovery would be nice but probably a lot more work.