Comparison of Birch vs other PPLs.

[thoth291]

Hello, folks and especially @lawmurray .

I stumbled upon this framework a few years ago while I was re-exploring PPL space in different environments: Python, Julia, C/C++ and even custom ones like Stan.

I'm not religious about what framework to use for my PPL projects.
So, every time I get a new project on the horizon - I use that as justification to dive into "something new" - to see if I should use something different this time for the project on my hands.

One thing which was totally missing in the docs around Birch is factual comparison with other PPLs (TF Probability, Pyro, NumPyro, Turing.jl, Stan).
It would be absolutely lovely to see:

• Functionality comparison
• Benchmarks for basic problems
• Some ways to integrate with ML frameworks (through bridges in Python, C++ or Julia).
  • Examples: Mixture Density Networks, Soft Decision Trees

Any plans to address those in the documentation?
Would really help me this time (third time) to think if I want to choose it for my work.