Aggregate Models, Not Explanations: Improving Feature Importance Estimation

ArXi:2602.11760v2 Announce Type: replace-cross Feature-importance methods show promise in transforming machine learning models from predictive engines into tools for scientific discovery. However, due to data sampling and algorithmic stochasticity, expressive models can be unstable, leading to inaccurate variable importance estimates and undermining their utility in critical biomedical applications.