Bayesian Force-Field Optimization

Structure and self-assembly are complex, emergent properties of matter that are often misrepresented by existing molecular simulation models. Bayesian optimization is an accurate and robust statistical method to optimize novel force fields based on experimental neutron/X-ray diffraction data to better model the structural behavior of liquid state systems.

We have developed an efficient and simple method to implement Bayesian inference for molecular simulations using local Gaussian process surrogate models [1] and are currently working on specific applications of Bayesian inference in neutron scattering.

[1] B.L. Shanks, H.W. Sullivan, A. R. Shazed and M.P. Hoepfner, Accelerated Bayesian Inference for Molecular Simulations using Local Gaussian Process Surrogate Models, J. Chem. Theory Comput., https://doi.org/10.1021/acs.jctc.3c01358