HDP priors?

[rsenne]

Heya @gdalle! Was wondering if you had been thinking about implementing the "Infinite Hidden Markov Model"? I was planning on adding this modeling structure to StateSpaceDynamics.jl soon, but since collaboration seems imminent, if this wasn't already a direction you had, I might try implementing this into HiddenMarkovModels.jl in the near-ish future.

[gdalle]

Hi @rsenne, thank you for the suggestion.
I had never read that paper before. I took a quick look and it seems to employ very different estimation algorithms (not exact but approximate), which means the overlap with HiddenMarkovModels.jl is rather low. I think this would make more sense in a separate package like InfiniteHiddenMarkovModels.jl, which could of course depend on mine for anything it needs.