Unsupervised neural controller for Reinforcement Learning action-selection: Learning to represent knowledge | IEEE Conference Publication | IEEE Xplore