Reproduce “Randomized Ablation Feature Importance” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Luke Merrick
The paper formalizes randomized ablation feature importance as the change in expected loss for a predictive model when one input is replaced by a fresh draw from its marginal distribution. It gives a finite-data estimator and confidence intervals for sampling and ablation uncertainty. The measure is model- and intervention-specific, not causal or conditional importance, and does not resolve correlated-feature effects.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.