Paths and consistency in additive cost sharing

75 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Axiomatic Attribution for Deep Networks (2017, background): “(For formal deﬁni-Axiomatic tions of the axioms and proof of Proposition 2, see Fried-man (Friedman, 2004).)”
Learning Explainable Models Using Attribution Priors (2019, methodology): “Since expected gradients is also a diagonal path method, it satisﬁes the same axioms as integrated gradients [10].”
Axiomatic cost and surplus sharing (2002, methodology): “(Friedman [1999]) i) A cost sharing method satisfies Dummy, Additivity and Demand Monotonicity if and only if it is an (infinite) convex combination of fixed path methods: [ (fixed path)]”
Improving performance of deep learning models with axiomatic attribution priors and expected gradients (2020, methodology): “Since EG is also a diagonal path method, it satisfies the same axioms as IG 44 .”
Feature relevance quantification in explainable AI: A causality problem (2019, methodology): “Integrated gradients can be generalized by integrating over an arbitrary path instead of the straight line. This attribution method is called path method and the following theorem holds. Theorem 1. ((Friedman, 2004, Theorem 1) and (Sundararajan et al., 2017, Theorem 1)) If an attribution method sati”