Behrens' Bayesian Learner Model : How to share parameters between steps?

[dddd1007]

@ricardoV94

Your code loses a lot of information between time steps by taking the mean of the posterior alone. For instance I don’t see how you can learn the volatility without sharing information across observations.

I read a PyMC3 tutorial Updating priors. I believe it gives me the solution which can save more information between time steps.