anykey.ai

Neuroevolution Learning

Watch a population of neural networks learn to navigate obstacles using a genetic algorithm. Over generations, the best agents reproduce, passing on their weights with slight mutations.

How it works

Generation

1

Alive

100

Best score

0

All-time

0

Best agent brain

How it works

  • Input: The bird sees its height, velocity, and the distance/gap of the next pipe.
  • Brain: A simple neural network (5 input, 8 hidden, 2 output nodes) processes the data.
  • Selection: Only birds that travel the furthest seed the next generation.
  • Mutation: Weights are slightly adjusted, allowing new strategies to emerge.