1
$\begingroup$

I want to use Viterbi algorithm, to decode an HMM sequence, but very few observations are missing in some of the steps or outliers. The hidden states in these steps are assumed to be the same as their previous states (or their adjacent states).

Can anyone tell me how can I set the emission probability in these steps. If I were to use Baum-Welch algorithm, what should I do.?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
3
$\begingroup$

It requires a bit of modification to the original Baum Welch. Shun-Zheng Yu and Hisashi Kobayashi wrote a paper about this, where they prove that their implementation of the EM algorithm for this problem of missing observations is correct and thereby produces a maximum likelihood.

Improve this answer
$\endgroup$

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.