1. Hidden Markov Models (HMMs) are used to model sequential data where the underlying process generating the observable outputs is not visible but assumed to be a Markov process with hidden states.
2. HMMs define transition probabilities between hidden states and emission probabilities of observable outputs for each state.
3. There are three typical problems for HMMs: likelihood computation, decoding the most likely sequence of hidden states, and learning the transition and emission probabilities from data.