Molecular Dynamics Simulations

Machine Learning for Uncertainty Quantification in Molecular Dynamics Simulations

Goal

In this project, the trainees will combine nonparametric uncertainty quantification techniques with data science methods, with the aim of advancing a new paradigm where molecular dynamics predictions are endowed with appropriate measures of confidence. The students will deploy and consolidate skills in data science, including surrogate modeling, manifold learning, and stochastic modeling, as well as in atomistic modeling and computational methods for materials science.

Image
Illustration of the integration of concepts and tools from data and materials science. Manifold learning, molecular dynamics, statistical modeling, ML-based surrogates.
Developing ML-based pipelines for predictive materials science under uncertainties