Less is more: sampling chemical space with active learning

By Justin S. Smith, B. Nebgen, Nicholas Lubbers, Olexandr Isayev, Adrian E. Roitberg. Journal of Chemical Physics, 2018

750 citations recorded by the source corpus.

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. 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.

Subfields: Computational Theory and Mathematics, Materials Chemistry, Physical and Theoretical Chemistry

View Paper: https://www.semanticscholar.org/paper/2a957bfcd11761cabb851e8a59fc4d022d92cd73

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Theory and Mathematics

In Materials Chemistry

In Physical and Theoretical Chemistry

Citations in the corpus

Machine-learned potentials for next-generation matter simulations (2021, methodology): “In Behler et al. 11 and Smith et al. 18 , different neural networks are trained for the chemical elements.”
The MLIP package: moment tensor potentials with MPI and active learning (2020, background): “It is the algorithm of choice for other neural-network-based potentials as well [70, 80].”
Approaching coupled cluster accuracy with a general-purpose neural network potential through transfer learning (2019, positive): “The existing ANI-1x active learning generated data set 25 is used to train an initial DFT (to the ω B97X/6-31G* model chemistry 63 ) potential, likewise dubbed ANI-1x.”
Human-in-the-loop machine learning: a state of the art (2022, background): “…find it at the basis of recommender systems (Rubens et al. 2015), for the construction of a reward estimation models (Lopes et al. 2009), and as the base for predicting molecular energetics (Smith et al. 2018) or predicting a heart disease (El-Hasnony et al. 2022) just to mention some of them.”
Deep Learning in Protein Structural Modeling and Design (2020, methodology): “extended this framework and tested its accuracy by simulating systems up to 312 atoms (Trp-cage) for 1 ns.(124,125) Another family that includes deep tensor neural networks(126) and SchNet(127) uses graph convolutions to learn a representation for each atom within its chemical environment.”