search

pymc-devs / pymc-experimental

https://pymc-experimental.readthedocs.io
Other
77 stars 49 forks source link

Specialized DiscreteMarkovChain step sampler #359

Closed ricardoV94 closed 2 weeks ago

ricardoV94 commented 2 months ago

Closes #324

It's just CategoricalGibbsMetropolis but accepts DiscreteMarkovChain and knows how to get the dimensionality of P.

Once we merge DMC into PyMC we can just extend the CategoricalGibbsMetropolis to accept it

ricardoV94 commented 2 months ago

Also I'm sure there must be more clever DMC samplers. This will update each step at a time, which I can imagine may be suboptimal. At least it updates in random order in every step

ricardoV94 commented 2 months ago

Seems like a fine stop-gap change

Maybe wait for the #361 to merge first then change where the test file is?

Sure