Hi, @ricardoV94. Is that Google Colab notebook you shared something you’re willing to make public? That was super useful.
I’m not dissatisfied with anything! I just think that with the goal of using Stan model coding style, it’ll be easier to just do that straight in JAX rather than going through PyMC. The Colab notebook clears up exactly how you can code models like Stan, which is super helpful.
The Wikipedia page just says this, and I still can’t tell what Theano does.
Theano is a Python library and optimizing compiler for manipulating and evaluating mathematical expressions, especially matrix-valued ones.[2] In Theano, computations are expressed using a NumPy-esque syntax and compiled to run efficiently on either CPU or GPU architectures.
It doesn’t even mention derivatives! In the past, I’ve seen theano described as a code transformation tool, but I think it may be more autodiff-based? The combination of not being able to even understand the top-level descriptions and it being in Python were dealbreakers for us.
I had no idea you could do that with Discourse. Thanks!
P.S. The Theano Wikipedia page is a mess. It mentions that Theano hit end of life and was forked by @fonnesbeck as Aeasara (no Wikipedia page) and then by the PyMC devs as PyTensor (also no Wikipedia page). I assume that means nobody is still maintaining Theano, even though the Theano page lists the PyMC devs. We struggled with getting a Wikipedia page for Stan, so I’m not surprised there’s not an Aesara or PyMC page, despite both projects being actively maintained.