I’m currently working on several Neural Networks using Elixir and Erlang’s OTP ecosystem. I recently just finished an example project called Deepnet. It’s one example of a 3 x 3 architecture. There are many more Network types I’m exploring. Here is a link to the post here for anyone interested in this topic.
Looks cool, what’s your end goal with this? I mean what’s your project goal.
I wouldn’t exactly call a 3x3 network “deep”…
That’s fair, Deep as in more than one Layer.
Right now, it’s about building different networks while testing their performance and finding limitations. Eventually, I’d like to do a major system written entirely in Elixir.
Ok sounds good. I would love to see more information about your journey over the time.
I took a look at the Matrix library you rely on… It’s a nested list! => this project has zero chance of ever being capable of learning real deep (& wide) network in a reasonable time. With that I mean network of thousands (if not hundreds of thousand or even millions) of input features and 3 or more layers , each with lesser features but still a lot. This matrix library can never efficiently handle such big matrices. See Big Data with Elixir for the same discussion at length.
So this can never be more than just a toy project.
Sorry to burst your bubble.
No bubble’s been burst here. I’m aware of the Matrix libraries limitations. That’s why in the post I mentioned Tensor that actually uses flow in the background instead of just using normal lists. That Matrix library was good enough for the problem set that was being illustrated in the writing.