Optimal allocation of computational resources based on Gaussian process: Application to molecular dynamics simulations

• Published in: Computational materials science 2021-02, Vol.188, p.110178, Article 110178
• Main Authors: Chilleri, John, He, Yanyan, Bedrov, Dmitry, Kirby, Robert M.
• Format: Article
• Language: English
• Subjects: Gaussian process; Glass-forming system; Molecular dynamics simulations; Optimal time allocation; Surrogate model; Uncertainty
• ISSN: 0927-0256; 1879-0801
• DOI: 10.1016/j.commatsci.2020.110178

[Display omitted] •A Gaussian process-based numerical optimization framework for optimal time allocation is proposed to guide the simulations over different locations.•A Gaussian process surrogate model is delivered under limited computational resources for molecular dynamics simulations.•Physical constraints such as non-negativity of diffusion coefficients are incorporated into the Gaussian process construction. Simulation models have been utilized in a wide range of real-world applications for behavior predictions of complex physical systems or material designs of large structures. While extensive simulation is mathematically preferable, external limitations such as available resources are often necessary considerations. With a fixed computational resource (i.e., total simulation time), we propose a Gaussian process-based numerical optimization framework for optimal time allocation over simulations at different locations, so that a surrogate model with uncertainty estimation can be constructed to approximate the full simulation. The proposed framework is demonstrated first via two synthetic problems, and later using a real test case of a glass-forming system with divergent dynamic relaxations where a Gaussian process is constructed to estimate the diffusivity and its uncertainty with respect to the temperature.