HMM Tutorial Series

I’ve been putting out a series of tutorials on hidden markov models that have generated quite a bit of interest and been featured on Towards Data Science. This overview page links to all the articles in one place and gives an overview of the series.

The articles so far in the series are:

• Hidden Markov Models: an Overview This article gives an overview of what HMMs are, the assumptions that go into them, and how you can analyze sequential data with the using the Viterbi algorithm.
• Training Hidden Markov Models This article discusses the somewhat subtle task of training a HMM on data for which you don’t have known state labels. In doing so it necessarily discusses issues like local optima and the EM algorithm.
• Continuous-time HMMs In HMMs the world is assumed to proceed in discrete timesteps, but many real situations proceed in real time, with irregular spacings between observations. This article generalizes the whole HMM discussion to this more general case.
• Continuous-time HMMs: When to Use Them: An article discussing CT-HMMs on more of a business/applications level, without unnecessary mathematical detail