Computationally Efficient Feature Significance and Importance for Machine Learning Models

2019/05/23 by Enguerrand Horel, Horel, Enguerrand, Kay Giesecke +1
Computer Science · Mathematics · #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine Learning and Algorithms #Machine Learning and Data Classification #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.1905.09849

openalex publication_date 2019/05/23 · openalex created_date 2019/05/29 · arxiv created 2019/10/12 · arxiv updated 2019/10/15 · openalex updated_date 2026/07/28

Abstract

We develop a simple and computationally efficient significance test for the features of a machine learning model. Our forward-selection approach applies to any model specification, learning task and variable type. The test is non-asymptotic, straightforward to implement, and does not require model refitting. It identifies the statistically significant features as well as feature interactions of any order in a hierarchical manner, and generates a model-free notion of feature importance. Experimental and empirical results illustrate its performance.

Citations

• The control of the false discovery rate in multiple testing under dependency
• How to Explain Individual Classification Decisions
• Axiomatic Attribution for Deep Networks
• Robust inference with knockoffs
• Sanity Checks for Saliency Maps
• Sensitivity based Neural Networks Explanations
• Regularization Matters: Generalization and Optimization of Neural Nets v.s. their Induced Kernel
• Understanding deep learning requires rethinking generalization
• A Survey Of Methods For Explaining Black Box Models
• On a Method to Measure Supervised Multiclass Model’s Interpretability: Application to Degradation Diagnosis (Short Paper)

Related

• Feature Importance Measure for Non-linear Learning Algorithms
• Understanding Global Feature Contributions With Additive Importance Measures
• Statistical Significance of Feature Importance Rankings
• Better Model Selection with a new Definition of Feature Importance
• CID: Measuring Feature Importance Through Counterfactual Distributions