# Hidden Markov Model - Estimating Transition and Emission CPDs from multiple sequences - not working

**URL:** <https://discourse.pymc.io/t/hidden-markov-model-estimating-transition-and-emission-cpds-from-multiple-sequences-not-working/6308>\
**Category:** Questions\
**Created:** [November 25, 2020, 3:52pm UTC](https://discourse.pymc.io/t/hidden-markov-model-estimating-transition-and-emission-cpds-from-multiple-sequences-not-working/6308 "2020-11-25T15:52:21Z")\
**Posts on this page:** 1\
**Showing post:** 2

<div class="post-metadata">

**Author:** ![junpenglao](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/junpenglao/32/8_2.png) [@junpenglao](https://discourse.pymc.io/u/junpenglao)\
**Post date:** [November 26, 2020, 9:41am UTC](https://discourse.pymc.io/t/hidden-markov-model-estimating-transition-and-emission-cpds-from-multiple-sequences-not-working/6308/2 "2020-11-26T09:41:40Z")

</div>

Strongly recommend you to try the marginalized HMM (see discussion in [How to marginalized Hidden Markov Model with categorical?](https://discourse.pymc.io/t/how-to-marginalized-hidden-markov-model-with-categorical/2230/9))

HMM model as in the main post use a discrete latent variable to represent the state, which cannot be sampled using HMC/NUTS - this usually means that you will get poor inference result. Marginalized HMM have other problems but a semi-supervised method should give reasonable inference result.

---

_[View the full topic](https://discourse.pymc.io/t/hidden-markov-model-estimating-transition-and-emission-cpds-from-multiple-sequences-not-working/6308)._