Generalizability of Machine Learning Models: Quantitative Evaluation of Three Methodological Pitfalls

no code implementations • 1 Feb 2022 • Farhad Maleki, Katie Ovens, Rajiv Gupta, Caroline Reinhold, Alan Spatz, Reza Forghani

We investigate three methodological pitfalls: (1) violation of independence assumption, (2) model evaluation with an inappropriate performance indicator or baseline for comparison, and (3) batch effect.

BIG-bench Machine Learning Data Augmentation +2