Recovering the Propensity Score from Biased Positive Unlabeled Data

• DOI: 10.1609/aaai.v36i6.20624
• Publication Date: 2022-07-04T11:12:35Z
• Authors: Walter Gerych; Thomas Hartvigsen; Luke Buquicchio; Emmanuel Agu; Elke Rundensteiner
• Keywords: Sample (material); Benchmark (surveying)

Positive-Unlabeled (PU) learning methods train a classifier to distinguish between the positive and negative classes given only unlabeled data. While traditional PU require labeled samples be an unbiased sample of distribution, in practice is often biased draw from true distribution. Prior work shows that if we know likelihood each instance will selected for labeling, referred as propensity score, then can used learning. Unfortunately, no prior has been proposed inference strategy which score identifiable. In this work, propose two sets assumptions under uniquely determined: one assumption made on functional form (requiring data distribution), second loosens while assuming score. We strategies case. Our empirical study our approach significantly outperforms state-of-the-art estimation rich variety benchmark datasets.