Martin,
Let me make this clear.
My Goals:
1. To benefit those of us who already chose Go and would like to embed ML into their software and would like module with a high-level API akin to Keras. There is a reason why Google adopted Keras to the official high-level API for Tensorflow.
My Non-Goals:
1. To convince anyone to start using Go in production.
2. To form a committee to decide on whether the world needs a new ML library.
3. To find out if I should continue working on Therfoo, that ship has sailed, I am already using it.
4. To debate whether I should port my software from Go to Python or to C
Any specific reason you are implementing
(correction, have implemented...) the Deep Learning module from Scratch?
admittedly I started writing this to develop my intuition as I studied ML and deep learning. I would presume it the same reason you started (and abandoned...?) https://github.com/chiteri/neural_networks_and_deep_learning ? There are a couple in existence already
(no there are not any production-grade deep learning libraries in Golang), probably well executed and tested. Perhaps you could join those projects and work on bug fixes and enhancements instead ....
Good luck in any case. thanks!
Kind regards,
David Ndungu