UFOs(well, they are not really flying) learning to drive.
UFOs are guided by a feedforward fully connected neural network whose weights are found using genetic algorithm.
Q-learning combined with a neural network as a function approximator. DQN algorithm is used.
The main goal of this project was not performance but to learn about Scala and functional programming in general. That's why you will find a neural network implementation from scratch where vectorization is not exploited.