隐马尔科夫模型(HMM)的基本问题.

UTF8gbsn

In the previous blog, we illustrated the definition of HMM and in this
blog we will list two important aspects of HMM.

The generation of observation sequence

Based on the definition of HMM, the generation of the observation
sequence O = { o 1 , o 2 , ⋯   , o T } O=\{o_1,o_2,\cdots, o_T\} O={o1​,o2​,⋯,oT​} is illustrated below:

• Input: HMM $\lambda =(A,B,\pi )$, and the length $T$ of observation
  sequence.

• output: Observation sequence O = { o 1 , o 2 , ⋯   , o T } O=\{o_1,o_2,\cdots, o_T\} O={o1​,o2​,⋯,oT​}

Algorithm:

1. The first state $i_1$ is generated by the initial state distribution
   $\pi$.

2. Let $t=1$

3. Based on the observation distribution $b_{i_t}(k)$ and $i_t$, we can
   get $o_t$ of state $i_t$.

4. Next, we can get our next state $i_{t+1}$ from the transition
   probability distribution { a i t , i t + 1 } \{a_{i_t,i_{t+1}}\} {ait​,it+1​​}.

5. Let $t=t+1$, if $t<T$, go back go step 3. if not, stop.

Three basic problems of HMM

1. After we defined the HMM $\lambda =(A,B,\pi )$ and generated the
   observation sequence $O=(o_1,o_2,\cdots ,o_T)$. There will be a
   problem of computing the conditional probability $P(O|\lambda )$.

2. Learning problem. Based on the current observation sequence
   O = { o 1 , o 2 , ⋯   , o T } O=\{o_1,o_2,\cdots, o_T\} O={o1​,o2​,⋯,oT​} how to estimate $\lambda =(A,B,\pi )$ that
   maximizes the probability of $P(O|\lambda )$.

3. Prediction. Based on the known HMM $\lambda =(A,B,\pi )$ and the
   observation sequence $O=(o_1,o_2,\cdots ,o_T)$, how to get the most
   likely state sequence $I=(i_1,i_2,\cdots ,i_T)$.