Null importance: Disentangling relevance for interpretable machine learning
A unified perspective based on null importance is developed: a population-level characterization of when a feature is irrelevant under a specified notion of relevance, which provides a common statistical language for relating scientific questions, data-generating assumptions, and algorithms, and clarify the conclusions...
Garvesh Raskutti, Kris Sankaran, Jia-Xin Ye
· 0 citations