Deep reinforcement learning for task-based feature learning in prosthetic vision | IEEE Conference Publication | IEEE Xplore