Polar Depth for Potentially Heavy-Tailed Data
Researchers have introduced a new statistical concept called Polar Depth, designed to analyze extreme values in multivariate heavy-tailed distributions. This novel approach, expressed in polar coordinates, offers a way to order and identify anomalies among extreme data points. The paper details the properties of Polar Depth and discusses its estimation, with numerical results demonstrating its effectiveness in anomaly detection. AI
IMPACT Introduces a new statistical tool for anomaly detection in heavy-tailed data, potentially improving AI model robustness.