HMM(几种期望)

UTF8gbsn

After we studied the HMM $\lambda =(A,B,\pi )$ and the backward and
forward methods of computing $P(O|\lambda )$, we should be able to
compute the following probabilities and expectations.

1. For a given $\lambda$ and $O$, the probability of
   ${\gamma }_t(i)=P(i_i=q_i|O,\lambda )$ is shown bellow.
   $${\gamma }_t(i)=P(i_t=q_i|O,\lambda )=\frac{P(i_t=q_i,O|\lambda )}{P(O|\lambda )}$$

   Based on the definition of ${\alpha }_t(i),{\beta }_t(i)$, we have :
   $${\alpha }_t(i){\beta }_t(i)=P(i_t=q_i,O|\lambda )$$

   Finally, we get:
   $${\gamma }_t(i)=\frac{{\alpha }_t(i){\beta }_t(i)}{P(O|\lambda )}=\frac{{\alpha }_t(i){\beta }_t(i)}{{\sum }_{j=1}^N{\alpha }_t(j){\beta }_t(j)}$$

2. For a given model $\lambda =(A,B,\pi )$ and observation sequence $O$,
   the probability of ${\xi }_t(i,j)=P(i_t=q_i,i_{t+1}=q_i|O,\lambda )$ is
   shown below.

   $${\xi }_t(i,j)=\frac{P(i_t=q_i,i_{t+1}=q_j,O|\lambda )}{P(O|\lambda )}=\frac{P(i_t=q_i,i_{t+1}=q_j,O|\lambda )}{{\sum }_{i=1}^N{\sum }_{j=1}^{N}P(i_t=q_i,i_{t+1}=q_j,O|\lambda )}$$

   where
   $P(i_t=q_i,i_{t+1}=q_j,O|\lambda )={\alpha }_t(i)a_{ij}{b}_j(o_{t+1}){\beta }_{t+1}(j)$.
   The original formula becomes

   $${\xi }_t(i,j)=\frac{{\alpha }_t(i)a_{ij}{b}_j(o_{t+1}){\beta }_{t+1}(j)}{{\sum }_{i=1}^N{\sum }_{j=1}^N{\alpha }_t(i)a_{ij}{b}_j(o_{t+1}){\beta }_{t+1}(j)}$$

3. The three expections:

   1. The expection of state $i$ with observation $O$ is:
      $$\sum _{t=1}^T{\gamma }_t(i)$$

   2. The expection of transfer from state $i$ with observation $O$
      is: $$\sum _{t=1}^{T-1}{\gamma }_t(i)$$

   3. The expection of transfer from state $i$ to state $j$ with
      observation $O$ is:

      $$\sum _{t=1}^{T-1}{\xi }_t(i,j)$$