NeuralTrex

I have always wanted to learn how to do machine learning, but had never done it before. I was inspired to create this by a video where a programmer taught a computer how to play Super Mario World. I chose google's T-Rex runner to teach the computer how to play.

I originally planned on using a neural network to teach it how to play, but found it to not be possible since it required supervised learning, which requires you to know that the output is to further train the model. Because of this, I decided to go with a Genetic Algorithm which uses unsupervised learning to find a solution.

This program was able to produce a player that achieved a score higher than 1000 points in just 9 short generations, which is better than most people who play the game.