reya: a project aiming to use a generalized reservoir computing architecture to play a wolfenstein-esque roguelike

I have disappeared for about a month, largely because I was busy adjusting to another bloody school year, but also because I spent most of my free time reading dozens (I may have surpassed a hundred) of cutting-edge research papers on neuromorphic computing, localized online learning, and reservoir computing. It was all very interesting, and I have concluded from them an idea for an architecture that I believe would allow for artificial intelligence with the capacity to learn complex behaviour, without much theoretical complexity (nor computational complexity, for that matter). I will call it reya, because I opened a book and that was the first word I saw.

I hope that reya, if it works, would blow all of my previous projects out of the water, but as of the time I am writing this, reya is only a terrible implementation of reservoir computing... written in Julia.

As much as I hate to use a language other than my beloved rust (it makes every other language feel badly designed), and even though debugging in julia is agnoizing, I have to admit that the rust linear algebra ecosystem, as rich as it is, is subpar if you want to do high-dimensional linear-algebra without headaches.

I might move back to rust after I prototype & test the architecture in julia, if the performance of julia turns out to be subpar.

reya repo: https://codeberg.org/0x177/reya