[QUESTION] How to get access to model parameters like initial/transition/emission matrix

[runfeng22]

I'm a beginner to HMM. Thanks for your effort on this amazing package. I fit data to the HMM model successfully, however, I want to ask how to get access to paramters like initial/transition/emission matrix.

I've found out that I can use self.starts/ends/edges to access some of it, but I didn't find how to get the emission matrix (or distributions in continuous data);

Another question is do you have any insights about why you use log value to describe the model rather than just the propablity?

Thank you!

[jmschrei]

self.edges should give you the transition matrix. self.distributions will give you a list of distribution objects, each of which will have their own parameters. See the distributions documentation for more information on that.

Log values are more numerically stable than original values and don't underflow when small probabilities are encountered, which is pretty frequently for large datasets/models.