Unsupervised State Representation Learning in Atari