Logical Characterizations of Recurrent Graph Neural Networks with Reals and Floats

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

5 Downloads (Pure)

Abstract

In pioneering work from 2019, Barceló and coauthors identified logics that precisely match the expressive power of constant iteration-depth graph neural networks (GNNs) relative to properties definable in first-order logic. In this article, we give exact logical characterizations of recurrent GNNs in two scenarios: (1) in the setting with floating-point numbers and (2) with reals. For floats, the formalism matching recurrent GNNs is a rule-based modal logic with counting, while for reals we use a suitable infinitary modal logic, also with counting. These results give exact matches between logics and GNNs in the recurrent setting without relativising to a background logic in either case, but using some natural assumptions about floating-point arithmetic. Applying our characterizations, we also prove that, relative to graph properties definable in monadic second-order logic (MSO), our infinitary and rule-based logics are equally expressive. This implies that recurrent GNNs with reals and floats have the same expressive power over MSO-definable properties and shows that, for such properties, also recurrent GNNs with reals are characterized by a (finitary!) rule-based modal logic. In the general case, in contrast, the expressive power with floats is weaker than with reals. In addition to logic-oriented results, we also characterize recurrent GNNs, with both reals and floats, via distributed automata, drawing links to distributed computing models.
Original languageEnglish
Title of host publicationThe Thirty-eighth Annual Conference on Neural Information Processing Systems
PublisherNeurIPS
Number of pages45
Publication statusE-pub ahead of print - 10 Dec 2024
Publication typeA4 Article in conference proceedings
EventConference on Neural Information Processing Systems - Vancouver, Canada
Duration: 10 Dec 202415 Dec 2024

Publication series

NameAdvances in neural information processing systems
ISSN (Print)1049-5258

Conference

ConferenceConference on Neural Information Processing Systems
Country/TerritoryCanada
CityVancouver
Period10/12/2415/12/24

Keywords

  • cs.LO
  • cs.AI
  • F.4.1; F.1.1; I.2.0

Publication forum classification

  • Publication forum level 3

Fingerprint

Dive into the research topics of 'Logical Characterizations of Recurrent Graph Neural Networks with Reals and Floats'. Together they form a unique fingerprint.

Cite this