A nonparametric two-sample test using a parametric integral probability metric
Researchers have introduced PReLU-TST, a novel nonparametric two-sample testing procedure designed to detect distributional differences between datasets. This new method utilizes a parametric integral probability metric (IPM) with a neural network discriminator, resulting in a test statistic named PReLU-IPM. Theoretical guarantees for PReLU-TST's consistency and asymptotic equivalence to existing IPM-based tests have been established. Empirical evaluations on simulated and real-world datasets indicate that PReLU-TST offers superior or comparable power to its competitors. AI
IMPACT Introduces a new statistical method for detecting distributional differences, potentially improving machine learning model evaluation.