Off-Policy Evaluation: Why Importance Sampling Gives Unbiased Estimates You Cannot Trust

Automated excerpt

Importance sampling gives an unbiased estimator of the new policy's value. Under one condition, this estimator is unbiased: the behaviour policy must have support everywhere the target policy does. Per-decision importance sampling exploits this: .

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.