One thing I wonder (maybe not a sensible question, I haven't really thought it through): A Kalman filter update is superficially like a HMM Forward Algorithm step, right? You have some belief, you have a new observation, and you have some new belief after.
Are there similar equivalences with other HMM algorithms for different questions? A Viterbi-like algorithm for the most likely path, for example, distinct from the series of Kalman-update argmaxes?
Are there similar equivalences with other HMM algorithms for different questions? A Viterbi-like algorithm for the most likely path, for example, distinct from the series of Kalman-update argmaxes?