Inverse Probability Weighting

A weighting method originating from importance sampling and causal inference. It weights the contribution of each sample by the inverse of its sampling probability so as to recover, in expectation, the result of full computation, thereby ensuring that the estimator is conditionally unbiased.