1 paper
Matthew Schlegel, Wesley Chung, Daniel Graves +2
Importance sampling (IS) is a common reweighting strategy for off-policy prediction in reinforcement learning. While it is consistent and unbiased, it can result in high variance u…