falsejam
A genetic algorithm slowly evolves to play Snake using random mutations of a neural network.
Generations 0 – 32
The top performing Snakes in each generation are used to breed the next.
Process Walkthrough: https://becominghuman.ai/designing-ai-solving-snake-with-evolution-f3dd6a9da867
Live interactive version: https://pbinggeser.github.io/snake-ai/
Github Repo: https://github.com/pbinggeser/snake-ai
Source
this is cool
Code? Very good approach and execution
I find the one on I2 in Generation 14 funny.
I3 if you counted from 1
What inputs do the snakes take?
Obviously next step would be to let all sneaks to run in the same large room and hit each other. This way you can breed even smarter sneaks
Genetic algorithm could be improved. Even at the end the player at the top right is continuously failing. Why was that gene not killed off?
That good after 24 generations, I just ran the interactive version and these dimwits can't figure out that hitting themselves is bad. Generation 87 and they still suck…
Can you please explain to me how you created a grid of canvas's, each with a separate game being played? anything would be super helpful!