skip to main content

DOE PAGESDOE PAGES

This content will become publicly available on May 22, 2019

Title: Less is more: Sampling chemical space with active learning

We present the development of accurate and transferable machine learning (ML) potentials for predicting molecular energetics is a challenging task. The process of data generation to train such ML potentials is a task neither well understood nor researched in detail. In this work, we present a fully automated approach for the generation of datasets with the intent of training universal ML potentials. It is based on the concept of active learning (AL) via Query by Committee (QBC), which uses the disagreement between an ensemble of ML potentials to infer the reliability of the ensemble’s prediction. QBC allows the presented AL algorithm to automatically sample regions of chemical space where the ML potential fails to accurately predict the potential energy. AL improves the overall fitness of ANAKIN-ME (ANI) deep learning potentials in rigorous test cases by mitigating human biases in deciding what new training data to use. AL also reduces the training set size to a fraction of the data required when using naive random sampling techniques. To provide validation of our AL approach, we develop the COmprehensive Machine-learning Potential (COMP6) benchmark (publicly available on GitHub) which contains a diverse set of organic molecules. Active learning-based ANI potentials outperform the originalmore » random sampled ANI-1 potential with only 10% of the data, while the final active learning-based model vastly outperforms ANI-1 on the COMP6 benchmark after training to only 25% of the data. Finally, we show that our proposed AL technique develops a universal ANI potential (ANI-1x) that provides accurate energy and force predictions on the entire COMP6 benchmark. Finally, this universal ML potential achieves a level of accuracy on par with the best ML potentials for single molecules or materials, while remaining applicable to the general class of organic molecules composed of the elements CHNO.« less
Authors:
ORCiD logo [1] ; ORCiD logo [2] ; ORCiD logo [2] ;  [3] ;  [1]
  1. Univ. of Florida, Gainesville, FL (United States)
  2. Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
  3. Univ. of North Carolina, Chapel Hill, NC (United States)
Publication Date:
Report Number(s):
LA-UR-18-30171
Journal ID: ISSN 0021-9606
Grant/Contract Number:
AC52-06NA25396
Type:
Accepted Manuscript
Journal Name:
Journal of Chemical Physics
Additional Journal Information:
Journal Volume: 148; Journal Issue: 24; Journal ID: ISSN 0021-9606
Publisher:
American Institute of Physics (AIP)
Research Org:
Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
Sponsoring Org:
USDOE Office of Science (SC); USDOE National Nuclear Security Administration (NNSA)
Country of Publication:
United States
Language:
English
Subject:
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CHEMISTRY; Material Science
OSTI Identifier:
1479911
Alternate Identifier(s):
OSTI ID: 1438295

Smith, Justin Steven, Nebgen, Benjamin Tyler, Lubbers, Nicholas Edward, Isayev, Olexandr, and Roitberg, Adrian E. Less is more: Sampling chemical space with active learning. United States: N. p., Web. doi:10.1063/1.5023802.
Smith, Justin Steven, Nebgen, Benjamin Tyler, Lubbers, Nicholas Edward, Isayev, Olexandr, & Roitberg, Adrian E. Less is more: Sampling chemical space with active learning. United States. doi:10.1063/1.5023802.
Smith, Justin Steven, Nebgen, Benjamin Tyler, Lubbers, Nicholas Edward, Isayev, Olexandr, and Roitberg, Adrian E. 2018. "Less is more: Sampling chemical space with active learning". United States. doi:10.1063/1.5023802.
@article{osti_1479911,
title = {Less is more: Sampling chemical space with active learning},
author = {Smith, Justin Steven and Nebgen, Benjamin Tyler and Lubbers, Nicholas Edward and Isayev, Olexandr and Roitberg, Adrian E},
abstractNote = {We present the development of accurate and transferable machine learning (ML) potentials for predicting molecular energetics is a challenging task. The process of data generation to train such ML potentials is a task neither well understood nor researched in detail. In this work, we present a fully automated approach for the generation of datasets with the intent of training universal ML potentials. It is based on the concept of active learning (AL) via Query by Committee (QBC), which uses the disagreement between an ensemble of ML potentials to infer the reliability of the ensemble’s prediction. QBC allows the presented AL algorithm to automatically sample regions of chemical space where the ML potential fails to accurately predict the potential energy. AL improves the overall fitness of ANAKIN-ME (ANI) deep learning potentials in rigorous test cases by mitigating human biases in deciding what new training data to use. AL also reduces the training set size to a fraction of the data required when using naive random sampling techniques. To provide validation of our AL approach, we develop the COmprehensive Machine-learning Potential (COMP6) benchmark (publicly available on GitHub) which contains a diverse set of organic molecules. Active learning-based ANI potentials outperform the original random sampled ANI-1 potential with only 10% of the data, while the final active learning-based model vastly outperforms ANI-1 on the COMP6 benchmark after training to only 25% of the data. Finally, we show that our proposed AL technique develops a universal ANI potential (ANI-1x) that provides accurate energy and force predictions on the entire COMP6 benchmark. Finally, this universal ML potential achieves a level of accuracy on par with the best ML potentials for single molecules or materials, while remaining applicable to the general class of organic molecules composed of the elements CHNO.},
doi = {10.1063/1.5023802},
journal = {Journal of Chemical Physics},
number = 24,
volume = 148,
place = {United States},
year = {2018},
month = {5}
}