This latest PR contains commits related to Bayesian model reduction. Once you've used a function like dirichlet_log_evidence (from the pymdp.maths module) to compare the free energies of a suite of models, you can use these reduction functions to selectively prune the observation and/or hidden state levels that receive low evidence/probability according to the model selected with BMR.
This latest PR contains commits related to Bayesian model reduction. Once you've used a function like
dirichlet_log_evidence(from thepymdp.mathsmodule) to compare the free energies of a suite of models, you can use these reduction functions to selectively prune the observation and/or hidden state levels that receive low evidence/probability according to the model selected with BMR.