Ab initio machine learning in chemical compound space

Huang, Bing; Lilienfeld, O. Anatole von

doi:10.48550/arxiv.2012.07502

Cited by 7 publications

(7 citation statements)

References 417 publications

(540 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Despite the good start with few training points, the curves develop quite slowly with increasing number of training data. The observed saturation is in contrast with the power law behavior that the learning curves should optimally follow with linear learning curves on the log-log scale [36,37]. One reason for the saturation could be the difficulty to resolve differences between some configurations as compared to others [38].…”

Section: Model Validation a Learning Curves And Cross Validationmentioning

confidence: 82%

Global optimization of atomic structures with gradient-enhanced Gaussian process regression

Kaappa,

del Río,

Jacobsen

2021

Preprint

View full text Add to dashboard Cite

Determination of atomic structures is a key challenge in the fields of computational physics and materials science, as a large variety of mechanical, chemical, electronic, and optical properties depend sensitively on structure. Here, we present a global optimization scheme where energy and force information from density functional theory (DFT) calculations is transferred to a probabilistic surrogate model to estimate both the potential energy surface (PES) and the associated uncertainties. The local minima in the surrogate PES are then used to guide the search for the global minimum in the DFT potential. We find that adding the gradients in most cases improves the efficiency of the search significantly. The method is applied to global optimization of [Ta2O5]x clusters with x = 1, 2, 3, and the surface structure of oxidized ZrN.

show abstract

Section: Model Validation a Learning Curves And Cross Validationmentioning

confidence: 82%

Global optimization of atomic structures with gradient-enhanced Gaussian process regression

Kaappa,

del Río,

Jacobsen

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The ANI-1 data set contains 2 × 10 7 structures of organic molecules. 211 For a summary of existing databases; see ref 212. With such approaches it may be possible to more broadly assess structural, substitutional, and electronic effects on chemical reactivity that undoubtedly are relevant for reaction outcomes.…”

Section: Organic Reactionsmentioning

confidence: 99%

Machine Learning for Chemical Reactions

2021

View full text Add to dashboard Cite

Machine learning (ML) techniques applied to chemical reactions have a long history. The present contribution discusses applications ranging from small molecule reaction dynamics to computational platforms for reaction planning. ML-based techniques can be particularly relevant for problems involving both computation and experiments. For one, Bayesian inference is a powerful approach to develop models consistent with knowledge from experiments. Second, ML-based methods can also be used to handle problems that are formally intractable using conventional approaches, such as exhaustive characterization of state-to-state information in reactive collisions. Finally, the explicit simulation of reactive networks as they occur in combustion has become possible using machine-learned neural network potentials. This review provides an overview of the questions that can and have been addressed using machine learning techniques, and an outlook discusses challenges in this diverse and stimulating field. It is concluded that ML applied to chemistry problems as practiced and conceived today has the potential to transform the way with which the field approaches problems involving chemical reactions, in both research and academic teaching.

show abstract

“…ML, in particular of solutions to the Schrodinger equation i.e. quantum machine learning [99][100][101][102][103] (QML), allows navigating chemical compound space (CCS) with high efficiency and precision. ML has become a popular avenue to material science with applications to atomization energies 104,105 , crystal formation energies 106 , carbenes 107 , excited states 108,109 , oxidation states 110 , nuclear magnetic resonance spectra 111 , reaction barriers 112,113 , magnetic systems 114 and charge transfer 115 or molecular fragments 116,117 .…”

Section: B Machine Learningmentioning

confidence: 99%

Ab initio machine learning of phase space averages

Weinreich,

Lemm,

von Rudorff

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Equilibrium structures determine material properties and biochemical functions. We propose to machine learn phase-space averages, conventionally obtained by ab initio or force-field based molecular dynamics (MD) or Monte Carlo simulations. In analogy to ab initio molecular dynamics (AIMD), our ab initio machine learning (AIML) model does not require bond topologies and therefore enables a general machine learning pathway to ensemble properties throughout chemical compound space. We demonstrate AIML for predicting Boltzmann averaged structures after training on hundreds of MD trajectories. AIML output is subsequently used to train machine learning models of free energies of solvation using experimental data, and reaching competitive prediction errors (MAE ∼ 0.8 kcal/mol) for out-of-sample molecules-within milli-seconds. As such, AIML effectively bypasses the need for MD or MC-based phase space sampling, enabling exploration campaigns throughout CCS at a much accelerated pace. We contextualize our findings by comparison to state-of-the-art methods resulting in a Pareto plot for the free energy of solvation predictions in terms of accuracy and time.

show abstract

Ab initio machine learning in chemical compound space

Cited by 7 publications

References 417 publications

Global optimization of atomic structures with gradient-enhanced Gaussian process regression

Global optimization of atomic structures with gradient-enhanced Gaussian process regression

Machine Learning for Chemical Reactions

Ab initio machine learning of phase space averages

Contact Info

Product

Resources

About