Playing Atari with Deep Reinforcement Learning

Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, and Martin Riedmiller submitted Playing Atari with Deep Reinforcement Learning on December 19, 2013. The paper presented a convolutional neural network trained with a variant of Q-learning to learn control policies directly from raw pixel input and evaluated the approach on seven Atari 2600 games from the Arcade Learning Environment.