Message-Passing Neural Network
Definition
An MPNN is a graph neural network that uses neighborhood messages to update node or graph states.
A class of graph neural networks that aggregate neighborhood messages and update node or graph states.
Simply explained
Like a group whose members combine information from their neighborhood and thereby update their view of the situation.
Why it matters in practice
Helps realistically assess model quality, limits, and technical decisions.
Example from business
A model predicts material properties from an atomic graph.
Limits and pitfalls
- Presenting MPNN as a single architecture or overlooking range and over-smoothing.
Origin & history
Gilmer and co-authors coined the MPNN framework for molecular properties in 2017.
Use in mid-sized companies
- Possible example: A specialist and development team models molecules as graphs and defines training and independent test data. They evaluate prediction quality and limits of neighborhood range.
Frequently asked questions
What does neural network with message passing mean?
A class of graph neural networks that aggregate neighborhood messages and update node or graph states.
Which mistakes should be avoided?
Presenting MPNN as a single architecture or overlooking range and over-smoothing.
Sources
Source checked on 10.10.2026. Editorial use cases are hypothetical; no evidence of real customer results.
Editorially reviewed on: