A feature is 'useful' if flipping its value can change a model's output; the paper proves this matches the logical notions of relevant and necessary features and analyzes the cost of computing them.
Compiling neural network classifiers into boolean circuits for efficient SHAP-score computation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Feature Relevancy, Necessity and Usefulness: Complexity and Algorithms
A feature is 'useful' if flipping its value can change a model's output; the paper proves this matches the logical notions of relevant and necessary features and analyzes the cost of computing them.