Exact Computation of Any-Order Shapley Interactions for Graph Neural Networks

Maximilian Muschalik, Fabian Fumagalli, Paolo Frazzetto, Janine Strotherm, Luca Hermes, Alessandro Sperduti, Eyke Hüllermeier, Barbara Hammer · 2025

Read Exact Computation of Any-Order Shapley Interactions for Graph Neural Networks by Maximilian Muschalik, Fabian Fumagalli, Paolo Frazzetto, Janine Strotherm, Luca Hermes, Alessandro Sperduti, Eyke Hüllermeier, Barbara Hammer (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-16944.txt.

Description

Maximilian Muschalik and seven co-authors compute exact Shapley interactions for graph neural networks, a sharper way to explain a single prediction. Their 2025 paper builds on Shapley values, which credit each input for a model's call. Single credits miss teamwork: in a graph, nodes sway each other, and group effects are the story.

The team show that the pattern of interactions survives the trip from node embeddings to the final graph call. That fact ties the cost of exact group credits to each node's reach, not to the whole graph. Their method, GraphSHAP-IQ, then computes group credits of any size exactly for common message-passing designs. On benchmarks it slashes the cost against brute force, and its approximations beat rival explainers. Maps of water networks and molecules show the credits at work.

Catalog record

CREATORMaximilian Muschalik, Fabian Fumagalli, Paolo Frazzetto, Janine Strotherm, Luca Hermes, Alessandro Sperduti, Eyke Hüllermeier, Barbara Hammer
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE107 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-16944

Browse the full library · About the Tower of Babel