Self-supervised graph attention networks for community-engaged lead contamination risk assessment

Researchers have developed a self-supervised graph attention network (SSGAT) to better predict lead contamination in residential water systems. The model uses spatial data to improve detection accuracy in data-limited urban environments.