Can the model state distributions be fixed?

[malonzm1]

Hi!

Can the model state distributions be fixed? Only fit the transition probability matrix?

Thanks and good day.

[jmschrei]

Yes. Pass frozen=True when constructing the distribution. If you're training a HMM you can also pass distribution_inertia=1.0 into the from_samples or fit methods.

[malonzm1]

Thanks. When I use frozen=True the state distribution parameters become NaN.

"parameters" : [
NaN,
NaN
],

But distribution_intertia=1.0 works. Thanks.

[jmschrei]

Glad that it's working.