Topology-Informed Machine Learning Improves Material Discovery

Strategic Insights -

Machine learning has shown considerable promise in designing novel materials, particularly in the design of glasses, but its efficacy is often hindered by limited material datasets. Researchers from the University of California Los Angeles (UCLA) have addressed this challenge by incorporating topological knowledge, derived from atomic structures, to inform machine learning models with physics-based insights. This approach, demonstrated through predicting Young's modulus of CaO-Al2O3-SiO2 glasses, has shown improved extrapolation abilities and promise for the discovery of new glass materials in unexplored domains.

Key Takeaways:

  • Researchers from the University of California Los Angeles (UCLA) have developed a topology-informed machine learning approach to improve material discovery.
  • The approach incorporates topological knowledge derived from atomic structures to inform machine learning models with physics-based insights.
  • The method demonstrated improved extrapolation abilities in predicting Young's modulus of CaO-Al2O3-SiO2 glasses.
  • The study has shown that the topology-informed approach can offer a more efficient and expedited pathway towards the discovery of new glass materials in unexplored domains.
  • The research team, led by Fabian Rosner, includes additional authors Kai Yang, Yu Song, Yuhai Li, Mathieu Bauchy, and Morten M. Smedskjaer.
  • The study has been peer-reviewed and published in the Journal of Non-crystalline Solids.

Statistics:

  • The study focused on predicting Young's modulus of CaO-Al2O3-SiO2 glasses.
  • The topology-informed machine learning approach demonstrated improved extrapolation abilities, maintaining comparable prediction accuracy within the training domain while significantly improving performance in extrapolating beyond the training domain.
  • The research was supported by the National Science Foundation (NSF).

Sources:

  • Enabling Extrapolation of Young's Modulus of Cao-AL2O3-SiO2 Ternary Glasses By Topology-informed Machine Learning. Journal of Non-crystalline Solids, 2025;666.
  • Journal of Non-crystalline Solids. Elsevier, Radarweg 29, 1043 NX Amsterdam, Netherlands.
  • NewsRx. Research Conducted at University of California Los Angeles (UCLA) Has Updated Our Knowledge about Machine Learning (Enabling Extrapolation of Young's Modulus of Cao-al2o3-sio2 Ternary Glasses By Topology-informed Machine Learning). Journal of Engineering. October 20, 2025; p 2998.