Skip to content

Playing around training neural networks with genetic algorithms

Notifications You must be signed in to change notification settings

LouieBloo/Neuroevolution

Repository files navigation

Neuroevolution

Playing around training neural networks with genetic algorithms. Currently have the retro game "Snake" built and have successfully trained a neural network using a standard GA approach (fitness,selection,crossover,mutation) to suruvie and score upwards 70+ food eaten.

Snake

20x20 grid, Touching walls or tail is death

  • Population: 100
  • Selection: Top 10%
  • Fitness function: # of food eaten
  • Mutation rate: 10%
  • Mutation amount: 0.1

About

Playing around training neural networks with genetic algorithms

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published