Multifidelity machine learning models for accurate bandgap predictions of solids
Here, we present a multifidelity cokriging statistical learning framework that combines variablefidelity quantum mechanical calculations of bandgaps to generate a machinelearned model that enables lowcost accurate predictions of the bandgaps at the highest fidelity level. Additionally, the adopted Gaussian process regression formulation allows us to predict the underlying uncertainties as a measure of our confidence in the predictions. In using a set of 600 elpasolite compounds as an example dataset and using semilocal and hybrid exchange correlation functionals within density functional theory as two levels of fidelities, we demonstrate the excellent learning performance of the method against actual high fidelity quantum mechanical calculations of the bandgaps. The presented statistical learning method is not restricted to bandgaps or electronic structure methods and extends the utility of high throughput property predictions in a significant way.
 Authors:

^{[1]}
;
^{[2]}
;
^{[2]}
 Los Alamos National Lab. (LANL), Los Alamos, NM (United States). Materials Science and Technology Division
 Los Alamos National Lab. (LANL), Los Alamos, NM (United States). Theoretical Division
 Publication Date:
 Report Number(s):
 LAUR1629228
Journal ID: ISSN 09270256; TRN: US1700513
 Grant/Contract Number:
 AC5206NA25396
 Type:
 Accepted Manuscript
 Journal Name:
 Computational Materials Science
 Additional Journal Information:
 Journal Volume: 129; Journal Issue: C; Journal ID: ISSN 09270256
 Publisher:
 Elsevier
 Research Org:
 Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
 Sponsoring Org:
 USDOE
 Country of Publication:
 United States
 Language:
 English
 Subject:
 36 MATERIALS SCIENCE; Double perovskites; Elpasolites; Materials informatics; Information fusion
 OSTI Identifier:
 1352377
 Alternate Identifier(s):
 OSTI ID: 1397619
Pilania, Ghanshyam, Gubernatis, James E., and Lookman, Turab. Multifidelity machine learning models for accurate bandgap predictions of solids. United States: N. p.,
Web. doi:10.1016/j.commatsci.2016.12.004.
Pilania, Ghanshyam, Gubernatis, James E., & Lookman, Turab. Multifidelity machine learning models for accurate bandgap predictions of solids. United States. doi:10.1016/j.commatsci.2016.12.004.
Pilania, Ghanshyam, Gubernatis, James E., and Lookman, Turab. 2016.
"Multifidelity machine learning models for accurate bandgap predictions of solids". United States.
doi:10.1016/j.commatsci.2016.12.004. https://www.osti.gov/servlets/purl/1352377.
@article{osti_1352377,
title = {Multifidelity machine learning models for accurate bandgap predictions of solids},
author = {Pilania, Ghanshyam and Gubernatis, James E. and Lookman, Turab},
abstractNote = {Here, we present a multifidelity cokriging statistical learning framework that combines variablefidelity quantum mechanical calculations of bandgaps to generate a machinelearned model that enables lowcost accurate predictions of the bandgaps at the highest fidelity level. Additionally, the adopted Gaussian process regression formulation allows us to predict the underlying uncertainties as a measure of our confidence in the predictions. In using a set of 600 elpasolite compounds as an example dataset and using semilocal and hybrid exchange correlation functionals within density functional theory as two levels of fidelities, we demonstrate the excellent learning performance of the method against actual high fidelity quantum mechanical calculations of the bandgaps. The presented statistical learning method is not restricted to bandgaps or electronic structure methods and extends the utility of high throughput property predictions in a significant way.},
doi = {10.1016/j.commatsci.2016.12.004},
journal = {Computational Materials Science},
number = C,
volume = 129,
place = {United States},
year = {2016},
month = {12}
}