I’ve done some experimentation with genetic algorithms and how to combine this with simple neural networks. My first attempt with genetic algorithms was a program that would generate a random color and would “learn” how to get to this color from any given random color. Of course there was no real learning and it used the principle of natural selection used by genetic algorithms to get closer and closer every generation. You can read more about genetic algorithms here. I’d later use this knowledge in the Genetic Robots project here.
Every flash in the GIF means that the current set of squares is filtered based on how “close that color is to the target”, only the best squares are kept and used for the next set.
The demo can be found here.
Press ENTER to change the target color.