UNATE: UNsupervised ATomic Embedding for crystal structures property prediction

Laura Solà-Garcia, Àlex Solé, Javier Ruiz-Hidalgo · 2026

Read UNATE: UNsupervised ATomic Embedding for crystal structures property prediction by Laura Solà-Garcia, Àlex Solé, Javier Ruiz-Hidalgo (2026) 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/arxiv-2605-25866.txt.

Description

Solà-Garcia, Solé and Ruiz-Hidalgo present UNATE, a framework learning atomic representations from unlabeled crystal structures. Crystal property prediction suffers from scarce labeled data and costly calculations. UNATE pairs a denoising autoencoder without supervision with contrastive learning that supervises itself, to build robust atomic features. Swapping atomic numbers for UNATE node embeddings beats the full-data baseline. Gains grow when labeled data is scarce, reaching about 10 percent with only a quarter of the labeled data.

Catalog record

CREATORLaura Solà-Garcia, Àlex Solé, Javier Ruiz-Hidalgo
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22880
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2605.25866

Browse the full library · About the Tower of Babel