I think this series of Jupyter notebook is really well done and helped me a lot learn when I started learning PyMC3.
3 Likes
Me too! also https://github.com/CamDavidsonPilon/Probabilistic-Programming-and-Bayesian-Methods-for-Hackers
However, both series are a bit dated now - PyMC3 evolve quite a lot since then, for example, we discourage the usage of pm.find_MAP() now.
We really need some new intro materials! I know @ericmjl is working on a new series: https://github.com/ericmjl/bayesian-analysis-recipes
1 Like
Yes really like Bayesian Methods for Hackers book.
Something missing from bayesian-analysis-recipes is a table of content of the available notebooks and what problem they are each trying to solve. I understand this is a work in progress, but I just tried opening each notebooks individually and it was kind of understand what it was about for some of them. But I learned something new from it where you can have multiple distributions with observed variables. I don`t know why I didn’t realize that before from all the other examples I read before…
Totally second the OP - the diagrams like this one reeeeally made stuff click in my head when I was first starting out:
2 Likes
Those diagrams should be used everywhere. Makes understanding bayesian stats so much easier.