https://compute.toys/view/342

re-implementation of znah's texture neural cellular automata in webgpu

can be viewed in chrome 113 or later
(given that you have a gpu)

I use the following command on Linux:

google-chrome-beta --enable-unsafe-webgpu --enable-features=Vulkan,UseSkiaRenderer

a fragment of live performance in secret place.
vcvrack gui stream is used as a texture input.
#webgl

реакция белоусова-жаботинского

This talk gave an inspiration for the creation of neural cellular automata model, and subsequently, the whole differentiable self-organizing system field.

To learn about neural cellular automata, start with these two papers on distill.pub

https://distill.pub/2020/growing-ca/

https://distill.pub/selforg/2021/textures/

this repository contains all relevant code for training such systems
https://github.com/google-research/self-organising-systems
some of the notebooks are covered in Alexander's tutorials on his youtube channel https://www.youtube.com/@zzznah/videos

on inference, NCA can be exported to webgl / webgpu environment, which allows them to run live in a browser.

https://www.shadertoy.com/view/cslcWM

https://compute.toys/view/342

compute.toys has vim mode out of the box

https://compute.toys/view/358
NCA minimum spanning tree, from ALIFE 2021 workshop

NCA learns translation operator