NN + RL for Flappy Birds

An agent trained to play Flappy Birds using Q learning and Neural Networks. Our project for the Neural Networks course in 2019. The learning environment used is from PyGame-Learning-Environment.



Results of our best agent ( vertical -> score, horizontal -> number of games played):

Two hidden layers, with leaky relu and tanh as activation functions and Adadelta as optimizer.