Neural Foundations · 2017
Neural Message Passing for Quantum Chemistry
Unified a family of graph neural network models into the message passing neural network (MPNN) framework, message, update and readout functions over graph edges, and showed it predicts quantum chemical properties on the QM9 benchmark to within chemical accuracy.
Editorial record
Plain-language summary
Several groups had proposed different ways to run neural networks over molecules and other graphs. This paper shows they are all instances of one scheme: each node collects messages from its neighbours, updates its state, and a readout function pools node states into a prediction. Framing it this way made the design space explicit and gave a strong, general recipe. MPNNs became the reference abstraction for graph learning and remain the standard vocabulary for describing graph neural networks.
Source record
Provenance
- Record ID
- P-675
- Record created
- 2026-09-26
- Last reviewed
- 2026-09-26
- Record version
- 1
- https://arxiv.org/abs/1704.01212
- arXiv:1704.01212
Citation caveat: Citation metadata is approximate and marked unverified in the source dataset.