Asymmetric error control under imperfect supervision: a label-noise-adjusted Neyman-Pearson umbrella algorithm

Label noise in data has long been an important problem supervised learning applications as it affects the effectiveness of many widely used classification methods. Recently, real-world applications, such medical diagnosis and cybersecurity, have generated renewed interest Neyman–Pearson (NP) paradigm, which constrains more severe type error (e.g., Type I error) under a preferred level while minimizing other II error). However, there little research on NP paradigm label noise. It is somewhat surprising that even when common classifiers ignore training stage, they are still able to control with high probability. price pay excessive conservativeness hence significant drop power (i.e., 1 - Assuming domain experts provide lower bounds corruption severity, we propose first theory-backed algorithm adapts most state-of-the-art methods paradigm. The resulting not only probability desired but also improve power.