Shapley Values with Uncertain Value Functions

被引：3

作者：

Heese, Raoul ^{[1
]}

Muecke, Sascha ^{[2
]}

Jakobs, Matthias ^{[2
]}

Gerlach, Thore ^{[3
]}

Piatkowski, Nico ^{[3
]}

机构：

[1] Fraunhofer ITWM, Kaiserslautern, Germany

[2] TU Dortmund, Dortmund, Germany

[3] Fraunhofer IAIS, St Augustin, Germany

来源：

ADVANCES IN INTELLIGENT DATA ANALYSIS XXI, IDA 2023 | 2023年 / 13876卷

关键词：

Shapley Values; Uncertainty; Explainable Machine Learning; Game Theory; CLASSIFICATIONS; GAME;

D O I：

10.1007/978-3-031-30047-9_13

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

We propose a novel definition of Shapley values with uncertain value functions based on first principles using probability theory. Such uncertain value functions can arise in the context of explainable machine learning as a result of non-deterministic algorithms. We show that random effects can in fact be absorbed into a Shapley value with a noiseless but shifted value function. Hence, Shapley values with uncertain value functions can be used in analogy to regular Shapley values. However, their reliable evaluation typically requires more computational effort.

引用

页码：156 / 168

页数：13

共 36 条

[1] Explaining individual predictions when features are dependent: More accurate approximations to Shapley values [J].