PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
October 17, 20250 citationsOpen Access

Logical Characterizations of GNNs with Mean Aggregation

View Full Paper
MSMoritz SchönherrCLCarsten Lutz

Key Points

  • Mean GNNs exhibit equal expressive power to ratio modal logic in a non-uniform context, enhancing understanding of graph representation.
  • In the uniform setting, mean GNNs maintain the same expressive power relative to MSO as alternation-free modal logic, showing its foundational role.
  • Mean aggregation's expressive power is notably less than sum and max aggregation GNNs under continuous combination functions and threshold classification.
  • When assumptions are relaxed, the expressive capabilities of mean GNNs increase significantly, prompting further exploration of aggregation effects.

Abstract

We study the expressive power of graph neural networks (GNNs) with mean as the aggregation function. In the non-uniform setting, we show that such GNNs have exactly the same expressive power as ratio modal logic, which has modal operators expressing that at least a certain ratio of the successors of a vertex satisfies a specified property. The non-uniform expressive power of mean GNNs is thus higher than that of GNNs with max aggregation, but lower than for sum aggregation--the latter are characterized by modal logic and graded modal logic, respectively. In the uniform setting, we show that the expressive power relative to MSO is exactly that of alternation-free modal logic, under the natural assumptions that combination functions are continuous and classification functions are thresholds. This implies that, relative to MSO and in the uniform setting, mean GNNs are strictly less expressive than sum GNNs and max GNNs. When any of the assumptions is dropped, the expressive power increases.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Schönherr et al. (2025) studied this question.

synapsesocial.com/papers/68f19f20de32064e504ddf10https://doi.org/10.48550/arxiv.2507.18145
Ask AI
Helpful
Bookmark
Share
View Full Paper

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Aggregate-Combine-Readout GNNs Are More Expressive Than Logic C22025
  2. 2Generalization Bounds for Message Passing Networks on Mixture of Graphons2024
  3. 3DGraFormer: Dynamic Graph Learning Guided Multi-Scale Transformer for Multivariate Time Series Forecasting2024
  4. 4A Logic for Reasoning About Aggregate-Combine Graph Neural Networks2024
  5. 5Are Targeted Messages More Effective?2024 · 1 citations