Tensorr / DoodleJumpAI

A Neural Network that learns to play Doodle Jump trained using a Genetic Algorithm

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

DoodleJumpAI

Download and Run

To run the program you will need Processing

Doodle

Neural Network

Each doodle contains a neural network. The neural network has an input layer of 5 neurons, 2 hidden layers of 4 neurons, and one output layer of 2 neurons.

Genetic Algorithm

A population of 200 doodles is created, each doodle with its own neural network. After all the doodles die some of the highest scoring ones are chosen to reproduce and create a new population of 200 doodles. This process is repeated for every generation.

Vision

The doodle can see in 5 directions. In each of these directions the doodle can see the distance to a pad, if there is one in that direction.

doodlevision

doodlejump-1

doodlejump-2

About

A Neural Network that learns to play Doodle Jump trained using a Genetic Algorithm

License:MIT License


Languages

Language:Processing 100.0%