The neural network based Δ-machine learning approach efficiently brings the DFT potential energy surface to the CCSD(T) quality: a case for the OH + CH<sub>3</sub>OH reaction

Song, Kai-Sheng; Li, Jun

doi:10.1039/d3cp00665d

Cited by 7 publications

(27 citation statements)

References 107 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…19 This method has been successfully applied in the HO 2 + HO 2 and OH + CH 3 OH reaction systems, in which only 14% and 5% of the configurations in the LL data set were needed, respectively. 19,24 These results indicate the efficiency and effectiveness of this NN-based Δ-ML approach in reducing computational costs while still achieving the accuracy of HL theory.…”

Section: Introductionmentioning

confidence: 68%

“…An alternative approach is to use the deviation between different NN fits of the correction PESs, which signals the sparsity of data, to identify configurations for HL calculations . This method has been successfully applied in the HO 2 + HO 2 and OH + CH 3 OH reaction systems, in which only 14% and 5% of the configurations in the LL data set were needed, respectively. , These results indicate the efficiency and effectiveness of this NN-based Δ-ML approach in reducing computational costs while still achieving the accuracy of HL theory.…”

Section: Introductionmentioning

confidence: 83%

“…Such a strategy is particularly useful for large polyatomic systems where CCSD(T) calculations are too expensive . With few exceptions, , however, these applications have mostly been restricted to nonreactive systems, which typically contain a smaller configuration space than reactive ones.…”

Section: Introductionmentioning

confidence: 99%

“…22 too expensive. 23 With few exceptions, 19,24 however, these applications have mostly been restricted to nonreactive systems, which typically contain a smaller configuration space than reactive ones. Unlike TL, which is based on the principle that knowledge acquired from solving one task (e.g., constructing a LL PES) is transferable to the new task (e.g., constructing a HL PES), Δ-ML focuses on correcting LL data with a few HL ones.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A Gaussian Process Based Δ-Machine Learning Approach to Reactive Potential Energy Surfaces

Liu,

Guo

2023

J. Phys. Chem. A

View full text Add to dashboard Cite

The Gaussian process (GP) is an efficient nonparametric machine learning (ML) method. A distinct advantage of the GP is its ability to provide an estimate of statistical uncertainties. This is particularly useful in constructing high-dimensional potential energy surfaces (PESs) from ab initio data as it offers an optimal way to add new geometries to reduce the overall error. In this work, GP is employed in the context of Δ-machine learning (Δ-ML), in which a correction PES to an inaccurate low-level PES is constructed using a small number of high-level ab initio calculations. This new method is tested in three prototypical reactive systems, namely, the H + H 2 → H + H 2 , OH + H 2 → H 2 O + H, and H + CH 4 → H 2 + CH 3 reactions. The results show that the GP-based Δ-ML approach is more efficient than its direct application in constructing high-level PESs. We also compare the new method to a previously proposed neural-network-based Δ-ML approach [Liu and Li J. Phys. Chem. Lett. 2022, 13, 4729−4738]. The results indicate that the two Δ-ML methods have comparable efficiencies in constructing accurate PESs.

show abstract

Section: Introductionmentioning

confidence: 68%

Section: Introductionmentioning

confidence: 83%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A Gaussian Process Based Δ-Machine Learning Approach to Reactive Potential Energy Surfaces

Liu,

Guo

2023

J. Phys. Chem. A

View full text Add to dashboard Cite

show abstract

“…In such cases, our proposed iterative methods of selection, namely "selection with random bins" and "selection with sorted bins", might outperform the "fully random" method. Additionally, active-learning methods 25 can also be used for selecting data points for high-level calculations. The potential advantages of these iterative methods in dealing with complex reaction systems are promising avenues for future research.…”

Section: Transfer Learning Data Selectionmentioning

confidence: 99%

Quasi-Classical Trajectory Calculation of Rate Constants Using an Ab Initio Trained Machine Learning Model (aML-MD) with Multifidelity Data

Shi,

Lele,

Jasper

et al. 2024

J. Phys. Chem. A

View full text Add to dashboard Cite

Machine learning (ML) provides a great opportunity for the construction of models with improved accuracy in classical molecular dynamics (MD). However, the accuracy of a ML trained model is limited by the quality and quantity of the training data. Generating large sets of accurate ab initio training data can require significant computational resources. Furthermore, inconsistent or incompatible data with different accuracies obtained using different methods may lead to biased or unreliable ML models that do not accurately represent the underlying physics. Recently, transfer learning showed its potential for avoiding these problems as well as for improving the accuracy, efficiency, and generalization of ML models using multifidelity data. In this work, ab initio trained ML-based MD (aML-MD) models are developed through transfer learning using DFT and multireference data from multiple sources with varying accuracy within the Deep Potential MD framework. The accuracy of the force field is demonstrated by calculating rate constants for the H + HO 2 → H 2 + 3 O 2 reaction using quasi-classical trajectories. We show that the aML-MD model with transfer learning can accurately predict the rate constants while reducing the computational cost by more than five times compared to the use of more expensive quantum chemistry training data sets. Hence, the aML-MD model with transfer learning shows great potential in using multifidelity data to reduce the computational cost involved in generating the training set for these potentials.

show abstract

Combining state-of-the-art quantum chemistry and machine learning make gold standard potential energy surfaces accessible for medium-sized molecules

Nandi,

Nagy

2024

Artificial Intelligence Chemistry

View full text Add to dashboard Cite

The neural network based Δ-machine learning approach efficiently brings the DFT potential energy surface to the CCSD(T) quality: a case for the OH + CH₃OH reaction

Abstract: The recently proposed permutationally invariant polynomial-neural network (PIP-NN) based ∆-machine learning (∆-ML) approach (PIP-NN ∆-ML, J. Phys. Chem. Lett. 2022, 13, 4729) is a flexible, general, and highly cost-efficient method...

Cited by 7 publications

References 107 publications

A Gaussian Process Based Δ-Machine Learning Approach to Reactive Potential Energy Surfaces

A Gaussian Process Based Δ-Machine Learning Approach to Reactive Potential Energy Surfaces

Quasi-Classical Trajectory Calculation of Rate Constants Using an Ab Initio Trained Machine Learning Model (aML-MD) with Multifidelity Data

Combining state-of-the-art quantum chemistry and machine learning make gold standard potential energy surfaces accessible for medium-sized molecules

Contact Info

Product

Resources

About

The neural network based Δ-machine learning approach efficiently brings the DFT potential energy surface to the CCSD(T) quality: a case for the OH + CH3OH reaction

Abstract: The recently proposed permutationally invariant polynomial-neural network (PIP-NN) based ∆-machine learning (∆-ML) approach (PIP-NN ∆-ML, J. Phys. Chem. Lett. 2022, 13, 4729) is a flexible, general, and highly cost-efficient method...

Cited by 7 publications

References 107 publications

A Gaussian Process Based Δ-Machine Learning Approach to Reactive Potential Energy Surfaces

A Gaussian Process Based Δ-Machine Learning Approach to Reactive Potential Energy Surfaces

Quasi-Classical Trajectory Calculation of Rate Constants Using an Ab Initio Trained Machine Learning Model (aML-MD) with Multifidelity Data

Combining state-of-the-art quantum chemistry and machine learning make gold standard potential energy surfaces accessible for medium-sized molecules

Contact Info

Product

Resources

About

The neural network based Δ-machine learning approach efficiently brings the DFT potential energy surface to the CCSD(T) quality: a case for the OH + CH₃OH reaction