What do Graph Neural Networks learn? Insights from Tropical Geometry

Tuan Anh Pham, Vikas Garg

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

Abstract

Graph neural networks (GNNs) have been analyzed from multiple perspectives, including the WL-hierarchy, which exposes limits on their expressivity to distinguish graphs. However, characterizing the class of functions that they learn has remained unresolved. We address this fundamental question for message passing GNNs under ReLU activations, i.e., the de-facto choice for most GNNs.We first show that such GNNs learn tropical rational signomial maps or continuous piecewise linear functions, establishing an equivalence with feedforward networks (FNNs). We then elucidate the role of the choice of aggregation and update functions, and derive the first general upper and lower bounds on the geometric complexity (i.e., the number of linear regions), establishing new results for popular architectures such as GraphSAGE and GIN. We also introduce and theoretically analyze several new architectures to illuminate the relative merits of the feedforward and the message passing layers, and the tradeoffs involving depth and number of trainable parameters. Finally, we also characterize the decision boundary for node and graph classification tasks.
Original languageEnglish
Title of host publicationAdvances in Neural Information Processing Systems 37 (NeurIPS 2024)
EditorsA. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak, C. Zhang
PublisherCurran Associates Inc.
ISBN (Print)9798331314385
Publication statusPublished - 2025
MoE publication typeA4 Conference publication
EventConference on Neural Information Processing Systems - Vancouver, Canada, Vancouver , Canada
Duration: 10 Dec 202415 Dec 2024
Conference number: 38
https://neurips.cc/Conferences/2024

Publication series

NameAdvances in Neural Information Processing Systems
PublisherCurran Associates, Inc.
Volume37
ISSN (Print)1049-5258

Conference

ConferenceConference on Neural Information Processing Systems
Abbreviated titleNeurIPS
Country/TerritoryCanada
CityVancouver
Period10/12/202415/12/2024
Internet address

Fingerprint

Dive into the research topics of 'What do Graph Neural Networks learn? Insights from Tropical Geometry'. Together they form a unique fingerprint.

Cite this