by Jason B. Gibson , Ajinkya C. Hire , Philip M. Dee, Oscar Barrera, Benjamin Geisler, Peter J. Hirschfeld, Richard G. Hennig • 23 days ago
Integrating deep learning with superconductivity research, the study introduces BETE-NET, a model that predicts the electron-phonon spectral function (α2F(ω)) using a database of 818 materials. This approach addresses the challenge of limited data by employing a bootstrapping technique to enhance predictions and reduce overfitting. The model significantly improves the search for high-temperature superconductors, demonstrating nearly five times the precision of random screening, thereby accelerating material discovery in this crucial field.