2017
DOI: 10.1039/c6sc05720a
|View full text |Cite
|
Sign up to set email alerts
|

ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost

Abstract: We demonstrate how a deep neural network (NN) trained on a data set of quantum mechanical (QM) DFT calculated energies can learn an accurate and transferable atomistic potential for organic molecules containing H, C, N, and O atoms.

Help me understand this report
View preprint versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

8
2,001
1
6

Year Published

2017
2017
2024
2024

Publication Types

Select...
5
3
2

Relationship

1
9

Authors

Journals

citations
Cited by 1,672 publications
(2,016 citation statements)
references
References 55 publications
8
2,001
1
6
Order By: Relevance
“…Recent work includes the general ANI-1 potential for the prediction of DFT energies [77] or a molecule specific parameterization of the internal energies for branched molecules with correlated dihedral potentials. Recent work includes the general ANI-1 potential for the prediction of DFT energies [77] or a molecule specific parameterization of the internal energies for branched molecules with correlated dihedral potentials.…”
Section: Morphologymentioning
confidence: 99%
“…Recent work includes the general ANI-1 potential for the prediction of DFT energies [77] or a molecule specific parameterization of the internal energies for branched molecules with correlated dihedral potentials. Recent work includes the general ANI-1 potential for the prediction of DFT energies [77] or a molecule specific parameterization of the internal energies for branched molecules with correlated dihedral potentials.…”
Section: Morphologymentioning
confidence: 99%
“…33 Neural networks have also been successfully applied in the cheminformatics domain through creative manipulations of 2D chemical structure and construction of the network architecture, 34–37 and as alternatives to computationally intensive quantum chemical calculations. 3840 …”
Section: Introductionmentioning
confidence: 99%
“…[12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27] They have been 3 used to discover materials [28][29][30][31][32][33][34][35][36][37] and study dynamical processes such as charge and exciton transfer. [38][39][40][41] Most related to this work are ML models of existing charge models, [9,[42][43][44] which are orders of magnitude faster than ab initio calculation.…”
Section: Molecular Size Training Datasetmentioning
confidence: 99%