When it comes to the training of a Hidden Markov Model using multiple training instances should I first take a single instance and train the model until the convergence and then move on to the next instance or should I use multiple instances sequentially to train the model?

  • 1
    $\begingroup$ Can you clarify what you mean by a single instance? From the way you put it, my understanding is that each instance is a single observation sequence, for instance a day's worth of observations. Is this what you mean? Also, I believe in the last sentence you mean simultaneously not sequentially. Is this correct? $\endgroup$
    – Zhubarb
    Oct 9, 2013 at 13:55

1 Answer 1


Trainig Hidden Markov Model using Maximum Likelihood criterion has two cases:

  1. Observed data: When your training data is fully observed and there is no hidden variables, the only thing you need to do is to count the frequencies and form the three matrices: $ \pi $,$A$ and $B$.
  2. Unobserved data: When there are hidden variables, there will be some free parameters which could not be found by counting. At this moment, Expectation-Maximization(EM) or namely Baum-Welch is used. In Baum-Welch algorithm there are two steps done on all instances in each training iteration: first is to calculate the expectation of training instances and second is to maximize it.

So, the answer to your question is in every training iteration, all instances are involved. Read this great tutorial on HMM, I'm sure it will help you.

  • $\begingroup$ What does "all instances are involved" mean? should they be concatenated and trained at once? or iterated over? $\endgroup$
    – Ran
    Apr 25, 2014 at 8:53
  • $\begingroup$ @Ran they are iterated over in each step of the training. sorry for the late reply I didn't notice the comment. $\endgroup$
    – Moh
    Jun 11, 2014 at 20:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.