Model selection in atomistic simulation
arXiv:2301.05287 · doi:10.1063/5.0142248
Abstract
There are many atomistic simulation methods with very different costs, accuracies, transferabilities, and numbers of empirical parameters. I show how statistical model selection can compare these methods fairly, even when they are very different. These comparisons are also useful for developing new methods that balance cost and accuracy. As an example, I build a semiempirical model for hydrogen clusters.
arXiv comment17 pages, 9 figures
References in corpus (9)
- Perspective on density functional theory
- Reconciling modern machine learning practice and the bias-variance trade-off
- Localization and delocalization errors in density functional theory and implications for band-gap prediction
- Recent developments in the PySCF program package
- Ground-state properties of the hydrogen chain: insulator-to-metal transition, dimerization, and magnetic phases
- Semiempirical Molecular Orbital Models based on the Neglect of Diatomic Differential Overlap Approximation
- Probabilistic performance estimators for computational chemistry methods: the empirical cumulative distribution function of absolute errors
- Trends in atomistic simulation software usage
- Semiempirical Hamiltonians learned from data can have accuracy comparable to Density Functional Theory