Direct gradient-based reinforcement learning | IEEE Conference Publication | IEEE Xplore