2022
DOI: 10.48550/arxiv.2204.12074
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

A structural optimization algorithm with stochastic forces and stresses

Abstract: We propose an algorithm for optimizations in which the gradients contain stochastic noise. This arises, for example, in structural optimizations when computations of forces and stresses rely on methods involving Monte Carlo sampling, such as quantum Monte Carlo or neural network states, or are performed on quantum devices which have intrinsic noise. Our proposed algorithm is based on the combination of two key ingredients: an update rule derived from the steepest descent method, and a staged scheduling of the … Show more

Help me understand this report
View published versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
1

Citation Types

0
1
0

Year Published

2022
2022
2022
2022

Publication Types

Select...
1

Relationship

0
1

Authors

Journals

citations
Cited by 1 publication
(1 citation statement)
references
References 31 publications
0
1
0
Order By: Relevance
“…50 An all-electron, localized-orbital ph-AFQMC calculation of the Fe(acac) 3 complex with around 1000 basis functions (cc-pVTZ) and a trial wavefunction with ∼100 determinants requires only 3 hours of wall-time, running on 100 nodes (on current Summit resources). Additional algorithmic advances include efficient ways to utilize multi-determinant trial wavefunctions, 48,51 a correlated sampling approach to more efficiently converge energy differences, 52 tensor decompositions, 53,54 frozen core and downfolding techniques, 55 stochastic resolutionof-the-identity strategies, 56 constraint release methods, nuclear gradients and geometry optimization, 57,58 and the back-propagation algorithm to estimate observables which do not commute with the Hamiltonian. 59 In the limit of an exact trial wavefunction, ph-AFQMC will recover the exact ground-state energy.…”
Section: Afqmc In Contextmentioning
confidence: 99%
“…50 An all-electron, localized-orbital ph-AFQMC calculation of the Fe(acac) 3 complex with around 1000 basis functions (cc-pVTZ) and a trial wavefunction with ∼100 determinants requires only 3 hours of wall-time, running on 100 nodes (on current Summit resources). Additional algorithmic advances include efficient ways to utilize multi-determinant trial wavefunctions, 48,51 a correlated sampling approach to more efficiently converge energy differences, 52 tensor decompositions, 53,54 frozen core and downfolding techniques, 55 stochastic resolutionof-the-identity strategies, 56 constraint release methods, nuclear gradients and geometry optimization, 57,58 and the back-propagation algorithm to estimate observables which do not commute with the Hamiltonian. 59 In the limit of an exact trial wavefunction, ph-AFQMC will recover the exact ground-state energy.…”
Section: Afqmc In Contextmentioning
confidence: 99%