Скачать или смотреть 100 generations of a pictures producing neuroevolutionary algorithm

The way this works:
-It creates a population of around 100 genomes that can be turned into neural networks that accept the x, y coordinates divided by the width or height of the picture, and that output the r, g, b and a components of a pixel.
-The user selects whichever pictures he likes enough, and when he chooses to step into the next generation, the genes of those pictures will be more likely to reproduce depending on the order in which he selected the pictures.
-The neuroevolutionary algorithm keeps different species in the population according to a certain measure of how different each genome is to the rest.
-If some species don't receive enough fitness, they die, which we can see through the video by the progressive disappearance of certain "kinds" of pictures.

Artículo relacionado: https://jonurenawriter.wordpress.com/...
Related article: https://jonurenawriter.wordpress.com/...

Video with full pictures of some of the genomes produced during this run:    • Pictures created during the uploaded neuro...