Hidden Semi-Markov Models

(2.17)

After that, let $t=t-{\hat{d}}_t$ and use Eqn (2.17) again, …, until $t\le 0$. Finally, a state sequence is obtained. However, it cannot be proved that this state sequence is the best.

2.3 Matrix Expression of the Forward–Backward Algorithm

One usually uses matrix-vector calculations to implement the forward–backward algorithms. To benefit from the matrix computation capability, it is better to express the forward–backward algorithms using matrices. To ...