Skip to content
[NeuroEvolution] Mini game where a Neural Network learns to avoid falling walls.
JavaScript HTML
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
lib
phrase
.gitignore
README.md
ga.js
game.js
index.html
player.js
wall.js

README.md

Evolution Simulation

This is a mini game I am currently working on to learn Genetic Algorithm.

DEMO

Objective

To avoid the falling walls.

As the score builds up, the velocity of the falling wall increases and the gap between the two walls decreases.

Neural Network Structure

3 - 5 - 1

Inputs to Neural Network

Three different parameters is fed into the Neural Netwok.

  • Height of the wall / Height of the game screen
  • End of the left wall / Width of the screen
  • Start of the seond wall / Width of the screen

Decision

There are 3 decisions to make.

  • Move Left if output < 0.33
  • Move Right if output < 0.67
  • Stay still
You can’t perform that action at this time.