sample-weighting

This paper proposes PUe, a framework for biased positive-unlabeled learning that uses normalized propensity scores and normalized inverse probability weighting to handle selection bias, improving classification under non-uniform label distributions.