SKEDSOFT

Learning hidden Markov models: Our final application of EM involves learning the transition probabilities in hidden Markov models (HMMs). Recall from Chapter 15 that a hidden Markov model can be represented by a dynamic Bayes net with a single discrete state variable, as illustrated in Figure 20.11. Each data point consists of an observation sequence of finite length, so the problem is to learn the transition probabilities from a set of observation sequences (or possibly from just one long sequence).
We have already worked out how to learn Bayes nets, but there is one complication: in Bayes nets, each parameter is distinct; in a hidden Markov model, on the other hand, the individual transition probabilities from state i to state j at time t, ijt =P(Xt 1 =jjXt =i), are repeated across time—that is, ijt =ij for all t. To estimate the transition probability from state i to state j, we simply calculate the expected proportion of times that the system undergoes a transition to state j when in state i:

Again, the expected counts are computed by any HMM inference algorithm. The forward– backward algorithm shown in Figure 15.4 can be modified very easily to compute the necessary probabilities. One important point is that the probabilities required are those obtained by smoothing rather than filtering; that is, we need to pay attention to subsequent evidence in estimating the probability that a particular transition occurred. As we said in Chapter 15, the evidence in a murder case is usually obtained after the crime (i.e., the transition from state i to state j) occurs.