Bounds for off-policy prediction in reinforcement learning | IEEE Conference Publication | IEEE Xplore