An enhanced least-squares approach for reinforcement learning | IEEE Conference Publication | IEEE Xplore