Yes, sorry, indeed I’ll switch over to pymc shortly.
Re: example, the x axis is the observation #. We want to update our inference after every observation. For example, at x=0, we have not observed any values and we just get the expected value of the latent. (this is for a coin-toss like paradigm, 0: tails; 1:heads).
As we observe more coin tosses (from the biased coin) we can update out prediction.
The red is indeed the Beta distribution with the update you mentioned. The blue is using pymc3 code above.
And I’d like to know why they are not nearly identical?