I just posted this in stackoverflow and the question was rejected given that was too broad. I am searching for answers to this, so, I help this community could help me:
I just started to learn PyMC3, and in general Probabilistic Programming, but after some exploration I have the folllowing question:
Obviously the first thing is that the language in the first 3 are different (Scala, Clojure, Python). The second difference is that (at least) from the examples around the web and books PyMC3 and Stan are most used in Bayesian Hierarchical Models/ Bayesian Inference and the examples (very few) from the former two are (apparently) more general.
In the book Practical probabilistic programming the author describes a
Probabilistic reasoning system and he describes it as capable of doing: (a) Predict future events, (b) Infer the cause of effects and © Learn from past events to better predict future events (pages 8-9), and he gives examples about this in the book using Figaro. Are PyMC3 and Stan capable of doing this? Are there examples of this? The same goes to Anglican, in several talks (including this one) they mention the use of Anglican for systems that shown the capabilities expressed above (in this case for a physical simulation in oil sea platforms).
Maybe a better way of asking this is possible to implement all the examples of the forementioned book in PyMC3 or Stan?