Building Action Sets in a Deep Reinforcement Learner | IEEE Conference Publication | IEEE Xplore