What kind of use cases are good for Nx?
Nx brings numerical computing power to the BEAM, so the Nx library and its friends are great if you need this sort of application on systems that were built on the BEAM.
One such use case, which also touches on the Nerves ecosystem, would be what we call Edge Computing – a small device, like a Raspberry Pi, could run an Elixir application
through Nerves which collects data from sensors and pre-processes that with Nx before sending to a central server.
Do you have a link to the GitHub?
This is the link to the repo which contains Nx, EXLA and Torchx: http://github.com/elixir-nx/nx
In the same organization you also find Axon (https://github.com/elixir-nx/axon) and Explorer (https://github.com/elixir-nx/explorer) which bring Neural Networks and Data Frames to the game, respectively.
What should developers know when getting started building neural networks with Nx?
We can divide the needed knowledge into two categories, one which pertains to neural networks themselves, and the other related to implementing them in Elixir.
For neural networks themselves, a basic understanding of statistics and linear algebra helps a lot, although lots of people start learning those subjects because of machine learning.
For writing them in Elixir, a basic understanding of the Elixir syntax is needed, and from there one can build up by learning the basics of how to use Nx, and then start using Axon, which is where you can actually define, train and use neural networks.
Paulo Valente on Twitter Paulo Valente (@polvalente) / Twitter