I have implemented a Dueling Double Deep Q-Network using TensorFlow + TFLearn as a course project. It can learn different Atari 2600 games from raw pixels and achieve a human-level success. Furthermore, it does all these with no adjustment to the learning algorithm or hyperparameters.

For those who have not heard about Deep Q-Networks (DQN), as DeepMind states it is the first demonstration of a general-purpose agent that is able to continually adapt its behavior without any human intervention, a major technical step forward in the quest for general AI.

To find out more, see my git repository.