Abstract by Chandramouli Nyshadham
Physics and Astronomy
Insights on materials space
In recent years, surrogate machine-learning models have transformed computational materials science by predicting properties of materials with the accuracy of quantum mechanics at a fraction of the cost. Using a kernel-based machine learning surrogate model, we present few insights on generating and choosing the training data for optimal modeling of materials space. Our insights are based on analyzing data from over 73,000 unrelaxed DFT calculations comprising 45 different materials and helped improve our model’s predictions by as much as 30% for some materials.