Hello everyone, just thought I’d share some resources that cropped up this past week about PyMC3’s and PyMC4’s design.
First, the PyMC developers got a paper accepted to a NeurIPS workshop! It’s about the design of PyMC4 around Python coroutines, and why we think this is the way forward for PyMC4. @ferrine will be presenting this work in Montreal, so drop by if you’ll be there!
Second, @junpenglao gave a talk at PyData Córdoba about the design of probabilistic programming frameworks, with a special focus on the design problems PyMC3 faced. Check out his slides here.
Finally, I published a new blog post about the same subject: what exactly does a probabilistic programming framework need to provide, and how are some of the more popular ones designed and structured? Check it out here: https://eigenfoo.xyz/prob-prog-frameworks/
Enjoy!