Neuroevolution of Neural Network of snakes in the Browser.
This is a demonstration of evolving a neural network thanks to genetics algorithms in the browser
using a multilayer perceptron (150-15-15-1).
The initial population contains 36 individuals, each assigned a different genome.
They will fight following a round-robin tournament.
At the end the top 7 are kept alive, and the remaining 29 are created by breeding from the 7.
Each snake has 50 sensors, each reporting 3 inputs:
- 1) The distance the sensor has hit something normalized between 0 and 1
- 2) 1 if this sensor touched the enemy body
- 3) 1 if this sensor touched the enemy body
If you reset the simulation, unselect "Show Curves" & "Show Debug" and put the simulation speed to something your CPU supports.
Code on Github