mlsanityData to 14 Oct 2026

Statistical Significance of Feature Importance Rankings

arXiv 2401.15800 ↗First posted 28 Jan 2024stat.ML cs.LGcited by 7

Not on the map. Why?

Abstract (arXiv)

Feature importance scores are ubiquitous tools for understanding the predictions of machine learning models. However, many popular attribution methods suffer from high instability due to random sampling. Leveraging novel ideas from hypothesis testing, we devise techniques that ensure the most important features are correct with high-probability guarantees. These assess the set of K top-ranked features, as well as the order of its elements. Given a set of local or global importance scores, we demonstrate how to retrospectively verify the stability of the highest ranks. We then introduce two efficient sampling algorithms that identify the K most important features, perhaps in order, with probability exceeding 1-α. The theoretical justification for these procedures is validated empirically on SHAP and LIME.

Key facts. Cited by 7 papers in the map (data through 14 Oct 2026). Builds on 24 of its 57 references that are in the map. First posted on arXiv 28 Jan 2024.

Builds on

24 of 57 references.

References · 24

  1. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  2. 2017A Unified Approach to Interpreting Model Predictionscited by 6,555
  3. 2016Distribution-Free Predictive Inference For Regressioncited by 668
  4. 2019Data Shapley: Equitable Valuation of Data for Machine Learningcited by 629
  5. 2018All Models are Wrong, but Many are Useful: Learning a Variable's Importance by Studying an Entire Class of Prediction Models Simultaneouslycited by 373
  6. 2019Explaining individual predictions when features are dependent: More accurate approximations to Shapley valuescited by 208
  7. 2020Improving KernelSHAP: Practical Shapley Value Estimation via Linear Regressioncited by 154
  8. 2019Asymmetric Shapley values: incorporating causal knowledge into model-agnostic explainabilitycited by 114
  9. 2021Sampling Permutations for Shapley Value Estimationcited by 102
  10. 2020Principles and Practice of Explainable Machine Learningcited by 99
  11. 2020Neuron Shapley: Discovering the Responsible Neuronscited by 98
  12. 2019Unrestricted Permutation forces Extrapolation: Variable Importance Requires at least One More Model, or There Is No Free Variable Importancecited by 87
12 more
  1. 2021GTG-Shapley: Efficient and Accurate Participant Contribution Evaluation in Federated Learningcited by 74
  2. 2019Veridical Data Sciencecited by 66
  3. 2021S-LIME: Stabilized-LIME for Model Explanationcited by 52
  4. 2021Shapley Counterfactual Credits for Multi-Agent Reinforcement Learningcited by 38
  5. 2019Global Aggregations of Local Explanations for Black Box modelscited by 30
  6. 2021The Shapley Value of Classifiers in Ensemble Gamescited by 23
  7. 2023Feature Importance: A Closer Look at Shapley Values and LOCOcited by 23
  8. 2016Rank Verification for Exponential Familiescited by 13
  9. 2023Stabilizing Estimates of Shapley Values with Control Variatescited by 9
  10. 2023Crime Prediction Using Machine Learning and Deep Learning: A Systematic Review and Future Directionscited by 8
  11. 2023Confident Feature Rankingcited by 6
  12. 2023SHAP@k:Efficient and Probably Approximately Correct (PAC) Identification of Top-k Featurescited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 3; 2026: 4.

7 papers.

  1. 2025Feature Relevancy, Necessity and Usefulness: Complexity and Algorithmscited by 1
  2. 2025Top-k Feature Importance Rankingcited by 1
  3. 2025Unifying Image Counterfactuals and Feature Attributions with Latent-Space Adversarial Attackscited by 0
  4. 2026Top-k on a Budget: Adaptive Ranking with Weak and Strong Oraclescited by 0
  5. 2026RoSHAP: A Distributional Framework and Robust Metric for Stable Feature Attributioncited by 0
  6. 2026K-IPO: Kendall-constrained Importance Preserving Oversampling for Imbalanced Tabular Datacited by 0
  7. 2026Measuring Explainer Stability via Attribution Separabilitycited by 0