From Garbage to Gold: A Data-Architectural Theory of Predictive Robustness

Terrence J. Lee-St. John, Jordan L. Lawson, Bartlomiej Piechowski-Jozwiak

Tabular machine learning presents a paradox: modern models achieve state-of-the-art performance using high-dimensional (high-D), collinear, error-prone data, defying the "Garbage I…