Shapley Values with Uncertain Value Functions
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.
- Published in:
International Symposium on Intelligent Data Analysis - Type:
Inproceedings - Authors:
Heese, Raoul; Mücke, Sascha; Jakobs, Matthias; Gerlach, Thore; Piatkowski, Nico - Year:
2023
Citation information
Heese, Raoul; Mücke, Sascha; Jakobs, Matthias; Gerlach, Thore; Piatkowski, Nico: Shapley Values with Uncertain Value Functions, International Symposium on Intelligent Data Analysis, 2023, https://link.springer.com/chapter/10.1007/978-3-031-30047-9_13, Heese.etal.2023a,
@Inproceedings{Heese.etal.2023a,
author={Heese, Raoul; Mücke, Sascha; Jakobs, Matthias; Gerlach, Thore; Piatkowski, Nico},
title={Shapley Values with Uncertain Value Functions},
booktitle={International Symposium on Intelligent Data Analysis},
url={https://link.springer.com/chapter/10.1007/978-3-031-30047-9_13},
year={2023},
abstract={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...}}
