Creating Gaussian process regression models for molecular simulations using adaptive sampling

Burn, Matthew J.; Popelier, Paul L. A.

doi:10.1063/5.0017887

“…As in NNs, a challenge arises from the absence of the reference atomic contributions because there is no unique way to calculate them using quantum chemistry. Nevertheless, quantum-chemistry approaches based on Bader analysis 97 have been applied to generate atomic contributions for training kernel methods 98 and NN 99 potentials.…”

Section: Kernel Methods Potentialsmentioning

confidence: 99%

“…As in NNs, a challenge arises from the absence of the reference atomic contributions because there is no unique way to calculate them using quantum chemistry. Nevertheless, quantum-chemistry approaches based on Bader analysis 97 have been applied to generate atomic contributions for training kernel methods 98 and NN 99 potentials. Alternatively, kernel methods can partition the total energy into atomic contribution during training by solving the correspondingly modified system of linear equations as is done in GAP, 42 FCHL, 44,95 and aSLATM.…”

Section: Machine Learning Algorithmsmentioning

confidence: 99%

Choosing the right molecular machine learning potential

Pinheiro

¹

,

Ge

²

,

Ferré

³

et al. 2021

View full text Add to dashboard Cite

show abstract

“…As in NNs, a challenge arises from the absence of the reference atomic contributions because there is no unique way to calculate them using quantum chemistry. Nevertheless, quantum-chemistry approaches based on Bader analysis 96 have been applied to generate atomic contributions for training kernel methods 97 and NN 98 potentials.…”

Section: Kernel Methods Potentialsmentioning

confidence: 99%

Choosing the right molecular machine learning potential

Pinheiro

¹

,

Ge

²

,

Ferré

³

et al. 2021

Preprint

View full text Add to dashboard Cite

Quantum-chemistry simulations based on potential energy surfaces of molecules provide invaluable insight into the physicochemical processes at the atomistic level and yield such important observables as reaction rates and spectra. Machine learning potentials promise to significantly reduce the computational cost and hence enable otherwise unfeasible simulations. However, the surging number of such potentials begs the question of which one to choose or whether we still need to develop yet another one. Here, we address this question by evaluating the performance of popular machine learning potentials in terms of accuracy and computational cost. In addition, we deliver structured information for non-specialists in machine learning to guide them through the maze of acronyms, recognize each potential's main features, and judge what they could expect from each one.

show abstract

“…As in NNs, a challenge arises from the absence of the reference atomic contributions because there is no unique way to calculate them using quantum chemistry. Nevertheless, quantumchemistry approaches based on Bader analysis 97 have been applied to generate atomic contributions for training kernel methods 98 and NN 99 potentials. Alternatively, kernel methods can partition the total energy into atomic contribution during training by solving the correspondingly modied system of linear equations as is done in GAP, 42 FCHL, 44,95 and aSLATM.…”

Section: Kernel Methods Potentialsmentioning

confidence: 99%

Choosing the right molecular machine learning potential

Pinheiro

¹

,

Ge

²

,

Ferré

³

et al. 2021

Preprint

0

View full text Add to dashboard Cite

Quantum-chemistry simulations based on potential energy surfaces of molecules provide invaluable insight into the physicochemical processes at the atomistic level and yield such important observables as reaction rates and spectra. Machine learning potentials promise to significantly reduce the computational cost and hence enable otherwise unfeasible simulations. However, the surging number of such potentials begs the question of which one to choose or whether we still need to develop yet another one. Here, we address this question by evaluating the performance of popular machine learning potentials in terms of accuracy and computational cost. In addition, we deliver structured information for non-specialists in machine learning to guide them through the maze of acronyms, recognize each potential's main features, and judge what they could expect from each one.

show abstract

Creating Gaussian process regression models for molecular simulations using adaptive sampling

Cited by 35 publications

References 58 publications

Choosing the right molecular machine learning potential

Choosing the right molecular machine learning potential

Choosing the right molecular machine learning potential

Choosing the right molecular machine learning potential

Contact Info

Product

Resources

About