[Question] How to retrieve dense transition matrix for a fitted hmm in version 1.0.4

[ko62147]

Hello,

I am fitting hidden Markov models (HMMs) to data using pomegranate version 1.0.4 and need to extract the dense (state) transition matrices of the HMMs. However, I am unable to find the method that returns the state transition matrix. I know that older versions of pomegranate (0.14.9) have a method for HMMs called dense_transition_matrix() that returns the transition probabilities between states. Is this method available in version 1.0.4 and how do I access it? Thanks for your anticipated reply.

[jmschrei]

self.edges should have the edges for both dense and sparse HMMs. https://github.com/jmschrei/pomegranate/blob/master/pomegranate/hmm/dense_hmm.py#L104

[ko62147]

Thanks for the reply. I noticed that self.edges only returns the state transition probabilities among distributions/states without the transitions from and to the silent (start and end) states. How would one go about including the start and end state transition probabilities in the matrix?

[jmschrei]

self.starts and self.ends contains the log probabilities.