This is my first post on this forum and it’s partially caused by the amusement and appreciation for Data Analysis and Machine Learning tools that Python devs has, like Tensorflow for example.
I know that historical Python and Erlang was adopted in different fields.
Until someone puts in the work to provide an API to the underlying math heavy routines written in C and C++ , it’s unlikely that there will be much available for the BEAM languages. There has been a lot of work in python to make these libraries accessible.
There is a python “port” for the BEAM, http://erlport.org/ that will help somewhat, but the kind of floating point mutable data computations that ML requires really don’t map to the BEAM very well. They don’t map well to python’s compute model well either, but python has a very good interface for working with external libraries.
There are improvements coming in the BEAM ( dirty schedulers) that will help change this situation, but Elixir is starting well behind in this area.