Critical Temperature Prediction for Superconducting Materials (PDF)

• Context: Graduate ML Course

• Overview: The project focused on developing a data-driven approach for predicting the critical temperature (Tc) of superconducting materials using machine learning techniques.

• Details: Deep Neural Network models (MLPs) were employed as surrogate models to learn the mapping between the features of known superconducting materials and their corresponding critical temperatures (Tc), as observed in laboratory settings. The study utilized a dataset of 21,263 superconductors, each described by 82 attributes. These attributes included key material characteristics such as the number of elements in their composition, average atomic weights, and the entropy of their atomic masses.