arXiv (Cornell University) · 2020 · 169 citations · 35 references
ComputingComputer ScienceData ScienceArtificial IntelligenceMachine LearningPredictive AnalyticsFeature AggregationFeature SelectionInteraction FeaturesNeural NetworksRepresentation LearningCross-feature Interaction ModelingFeature ImportanceImportance MeasuresInterpretabilityModel InterpretabilityModel-agnostic Explanation
Understanding the inner workings of complex machine learning models is a long-standing problem and most recent research has focused on local interpretability. To assess the role of individual input features in a global sense, we explore the perspective of defining feature importance through the predictive power associated with each feature. We introduce two notions of predictive power (model-based and universal) and formalize this approach with a framework of additive importance measures, which unifies numerous methods in the literature. We then propose SAGE, a model-agnostic method that quantifies predictive power while accounting for feature interactions. Our experiments show that SAGE can be calculated efficiently and that it assigns more accurate importance values than other methods.
35
Leo Breiman · Machine Learning · 2001 · 126.2K citations · Full text
UCI Machine Learning Repository
Arthur Asuncion · Medical Entomology and Zoology · 2007 · 24.4K citations
Marco Túlio Ribeiro, Sameer Singh, Carlos Guestrin · 2016 · 15.2K citations
Gene Selection for Cancer Classification using Support Vector Machines
Isabelle Guyon, Jason Weston, S. Barnhill et al. · Machine Learning · 2002 · 9.8K citations · Full text