Capsule graph networks for accurate and interpretable crystalline materials property prediction
Eddah K Sure, Xing Wu, Quan Qian · 2025
Read Capsule graph networks for accurate and interpretable crystalline materials property prediction by Eddah K Sure, Xing Wu, Quan Qian (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-12865943.txt.
Description
Eddah Sure and co-authors teach a neural network to see crystals the way crystals are built: parts inside wholes. Graph networks predict material properties well but explain little and ignore the hierarchy from atoms to polyhedra to extended motifs. Their capsule graph network, equipped with symmetry-respecting message passing, organizes what it learns into capsules that stand for local environments.
On standard data sets the model beat a leading crystal graph network at formation energy and matched another at band gap. The capsules themselves became readable: routing by agreement gathers environments like titanium-oxygen octahedra into higher-level units tied to the property. The authors present the design as prediction with a built-in explanation, an unsupervised route to discovering the motifs that matter.
Catalog record
| CREATOR | Eddah K Sure, Xing Wu, Quan Qian |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |