Analyzing machine learning models to accelerate generation of fundamental materials insights
Abstract
Abstract Machine learning for materials science envisions the acceleration of basic science research through automated identification of key data relationships to augment human interpretation and gain scientific understanding. A primary role of scientists is extraction of fundamental knowledge from data, and we demonstrate that this extraction can be accelerated using neural networks via analysis of the trained data model itself rather than its application as a prediction tool. Convolutional neural networks excel at modeling complex data relationships in multi-dimensional parameter spaces, such as that mapped by a combinatorial materials science experiment. Measuring a performance metric in a given materials space provides direct information about (locally) optimal materials but not the underlying materials science that gives rise to the variation in performance. By building a model that predicts performance (in this case photoelectrochemical power generation of a solar fuels photoanode) from materials parameters (in this case composition and Raman signal), subsequent analysis of gradients in the trained model reveals key data relationships that are not readily identified by human inspection or traditional statistical analyses. Human interpretation of these key relationships produces the desired fundamental understanding, demonstrating a framework in which machine learning accelerates data interpretation by leveraging the expertize of themore »
- Authors:
- Publication Date:
- Research Org.:
- California Institute of Technology (CalTech), Pasadena, CA (United States). Joint Center for Artificial Photosynthesis (JCAP)
- Sponsoring Org.:
- USDOE Office of Science (SC), Basic Energy Sciences (BES)
- OSTI Identifier:
- 1619633
- Alternate Identifier(s):
- OSTI ID: 1575062
- Grant/Contract Number:
- SC0004993
- Resource Type:
- Published Article
- Journal Name:
- npj Computational Materials
- Additional Journal Information:
- Journal Name: npj Computational Materials Journal Volume: 5 Journal Issue: 1; Journal ID: ISSN 2057-3960
- Publisher:
- Nature Publishing Group
- Country of Publication:
- United Kingdom
- Language:
- English
- Subject:
- 36 MATERIALS SCIENCE
Citation Formats
Umehara, Mitsutaro, Stein, Helge S., Guevarra, Dan, Newhouse, Paul F., Boyd, David A., and Gregoire, John M. Analyzing machine learning models to accelerate generation of fundamental materials insights. United Kingdom: N. p., 2019.
Web. doi:10.1038/s41524-019-0172-5.
Umehara, Mitsutaro, Stein, Helge S., Guevarra, Dan, Newhouse, Paul F., Boyd, David A., & Gregoire, John M. Analyzing machine learning models to accelerate generation of fundamental materials insights. United Kingdom. https://doi.org/10.1038/s41524-019-0172-5
Umehara, Mitsutaro, Stein, Helge S., Guevarra, Dan, Newhouse, Paul F., Boyd, David A., and Gregoire, John M. Fri .
"Analyzing machine learning models to accelerate generation of fundamental materials insights". United Kingdom. https://doi.org/10.1038/s41524-019-0172-5.
@article{osti_1619633,
title = {Analyzing machine learning models to accelerate generation of fundamental materials insights},
author = {Umehara, Mitsutaro and Stein, Helge S. and Guevarra, Dan and Newhouse, Paul F. and Boyd, David A. and Gregoire, John M.},
abstractNote = {Abstract Machine learning for materials science envisions the acceleration of basic science research through automated identification of key data relationships to augment human interpretation and gain scientific understanding. A primary role of scientists is extraction of fundamental knowledge from data, and we demonstrate that this extraction can be accelerated using neural networks via analysis of the trained data model itself rather than its application as a prediction tool. Convolutional neural networks excel at modeling complex data relationships in multi-dimensional parameter spaces, such as that mapped by a combinatorial materials science experiment. Measuring a performance metric in a given materials space provides direct information about (locally) optimal materials but not the underlying materials science that gives rise to the variation in performance. By building a model that predicts performance (in this case photoelectrochemical power generation of a solar fuels photoanode) from materials parameters (in this case composition and Raman signal), subsequent analysis of gradients in the trained model reveals key data relationships that are not readily identified by human inspection or traditional statistical analyses. Human interpretation of these key relationships produces the desired fundamental understanding, demonstrating a framework in which machine learning accelerates data interpretation by leveraging the expertize of the human scientist. We also demonstrate the use of neural network gradient analysis to automate prediction of the directions in parameter space, such as the addition of specific alloying elements, that may increase performance by moving beyond the confines of existing data.},
doi = {10.1038/s41524-019-0172-5},
journal = {npj Computational Materials},
number = 1,
volume = 5,
place = {United Kingdom},
year = {Fri Mar 08 00:00:00 EST 2019},
month = {Fri Mar 08 00:00:00 EST 2019}
}
https://doi.org/10.1038/s41524-019-0172-5
Web of Science
Works referenced in this record:
A simple constrained machine learning model for predicting high-pressure-hydrogen-compressor materials
journal, January 2018
- Hattrick-Simpers, Jason R.; Choudhary, Kamal; Corgnale, Claudio
- Molecular Systems Design & Engineering, Vol. 3, Issue 3
Learning atoms for materials discovery
journal, June 2018
- Zhou, Quan; Tang, Peizhe; Liu, Shenxiu
- Proceedings of the National Academy of Sciences, Vol. 115, Issue 28
Effects of molybdenum substitution on the photocatalytic behavior of BiVO4
journal, January 2008
- Yao, Weifeng; Iwai, Hideo; Ye, Jinhua
- Dalton Transactions, Issue 11
Systematic behaviour in trivalent lanthanide charge transfer energies
journal, November 2003
- Dorenbos, P.
- Journal of Physics: Condensed Matter, Vol. 15, Issue 49
Transparent, explainable, and accountable AI for robotics
journal, May 2017
- Wachter, Sandra; Mittelstadt, Brent; Floridi, Luciano
- Science Robotics, Vol. 2, Issue 6
Identification of a Blue Photoluminescent Composite Material from a Combinatorial Library
journal, March 1998
- Wang, J.
- Science, Vol. 279, Issue 5357
Machine learning modeling of superconducting critical temperature
journal, June 2018
- Stanev, Valentin; Oses, Corey; Kusne, A. Gilad
- npj Computational Materials, Vol. 4, Issue 1
Combinatorial Electrochemistry: A Highly Parallel, Optical Screening Method for Discovery of Better Electrocatalysts
journal, June 1998
- Reddington, E.
- Science, Vol. 280, Issue 5370
High-throughput determination of structural phase diagram and constituent phases using GRENDEL
journal, October 2015
- Kusne, A. G.; Keller, D.; Anderson, A.
- Nanotechnology, Vol. 26, Issue 44
Accelerated search for materials with targeted properties by adaptive design
journal, April 2016
- Xue, Dezhen; Balachandran, Prasanna V.; Hogden, John
- Nature Communications, Vol. 7, Issue 1
A Universal 3D Voxel Descriptor for Solid-State Material Informatics with Deep Convolutional Neural Networks
journal, December 2017
- Kajita, Seiji; Ohba, Nobuko; Jinnouchi, Ryosuke
- Scientific Reports, Vol. 7, Issue 1
Combinatorial Materials Sciences: Experimental Strategies for Accelerated Knowledge Discovery
journal, August 2008
- Rajan, Krishna
- Annual Review of Materials Research, Vol. 38, Issue 1
Perspective: Composition–structure–property mapping in high-throughput experiments: Turning data into knowledge
journal, May 2016
- Hattrick-Simpers, Jason R.; Gregoire, John M.; Kusne, A. Gilad
- APL Materials, Vol. 4, Issue 5
Deep Learning of Atomically Resolved Scanning Transmission Electron Microscopy Images: Chemical Identification and Tracking Local Transformations
journal, October 2016
- Ziatdinov, Maxim; Dyck, Ondrej; Maksov, Artem
- ACS Nano, Vol. 11, Issue 12
Solar fuels photoanode materials discovery by integrating high-throughput theory and experiment
journal, March 2017
- Yan, Qimin; Yu, Jie; Suram, Santosh K.
- Proceedings of the National Academy of Sciences, Vol. 114, Issue 12
Building data-driven models with microstructural images: Generalization and interpretability
journal, December 2017
- Ling, Julia; Hutchinson, Maxwell; Antono, Erin
- Materials Discovery, Vol. 10
Mastering the game of Go with deep neural networks and tree search
journal, January 2016
- Silver, David; Huang, Aja; Maddison, Chris J.
- Nature, Vol. 529, Issue 7587
Data Mined Ionic Substitutions for the Discovery of New Compounds
journal, January 2011
- Hautier, Geoffroy; Fischer, Chris; Ehrlacher, Virginie
- Inorganic Chemistry, Vol. 50, Issue 2
Unraveling compositional effects on the light-induced oxygen evolution in Bi(V–Mo–X)O 4 material libraries
journal, January 2017
- Gutkowski, R.; Khare, C.; Conzuelo, F.
- Energy & Environmental Science, Vol. 10, Issue 5
Structural, electronic and optical features of molybdenum-doped bismuth vanadium oxide
journal, March 2015
- Merupo, V. I.; Velumani, S.; Oza, G.
- Materials Science in Semiconductor Processing, Vol. 31
Learning surface molecular structures via machine vision
journal, August 2017
- Ziatdinov, Maxim; Maksov, Artem; Kalinin, Sergei V.
- npj Computational Materials, Vol. 3, Issue 1
Vanadium(V) environments in bismuth vanadates: A structural investigation using Raman spectroscopy and solid state 51V NMR
journal, February 1991
- Hardcastle, Franklin D.; Wachs, Israel E.; Eckert, Hellmut
- Journal of Solid State Chemistry, Vol. 90, Issue 2
Towards fully autonomous driving: Systems and algorithms
conference, June 2011
- Levinson, Jesse; Askeland, Jake; Becker, Jan
- 2011 IEEE Intelligent Vehicles Symposium (IV)
Microstructure recognition using convolutional neural networks for prediction of ionic conductivity in ceramics
journal, December 2017
- Kondo, Ruho; Yamakawa, Shunsuke; Masuoka, Yumi
- Acta Materialia, Vol. 141
Low-level interpretability and high-level interpretability: a unified view of data-driven interpretable fuzzy system modelling
journal, December 2008
- Zhou, Shang-Ming; Gan, John Q.
- Fuzzy Sets and Systems, Vol. 159, Issue 23
Transferability in Machine Learning for Electronic Structure via the Molecular Orbital Basis
journal, July 2018
- Welborn, Matthew; Cheng, Lixue; Miller, Thomas F.
- Journal of Chemical Theory and Computation, Vol. 14, Issue 9
The 2019 materials by design roadmap
journal, October 2018
- Alberi, Kirstin; Nardelli, Marco Buongiorno; Zakutayev, Andriy
- Journal of Physics D: Applied Physics, Vol. 52, Issue 1
Discovery of a useful thin-film dielectric using a composition-spread approach
journal, March 1998
- van Dover, R. B.; Schneemeyer, L. F.; Fleming, R. M.
- Nature, Vol. 392, Issue 6672
Autonomy in materials research: a case study in carbon nanotube growth
journal, October 2016
- Nikolaev, Pavel; Hooper, Daylond; Webber, Frederick
- npj Computational Materials, Vol. 2, Issue 1
Applications of high throughput (combinatorial) methodologies to electronic, magnetic, optical, and energy-related materials
journal, June 2013
- Green, Martin L.; Takeuchi, Ichiro; Hattrick-Simpers, Jason R.
- Journal of Applied Physics, Vol. 113, Issue 23
Automated Phase Mapping with AgileFD and its Application to Light Absorber Discovery in the V–Mn–Nb Oxide System
journal, December 2016
- Suram, Santosh K.; Xue, Yexiang; Bai, Junwen
- ACS Combinatorial Science, Vol. 19, Issue 1
A general-purpose machine learning framework for predicting properties of inorganic materials
journal, August 2016
- Ward, Logan; Agrawal, Ankit; Choudhary, Alok
- npj Computational Materials, Vol. 2, Issue 1
Multi-modal optimization of bismuth vanadate photoanodes via combinatorial alloying and hydrogen processing
journal, January 2019
- Newhouse, P. F.; Guevarra, D.; Umehara, M.
- Chemical Communications, Vol. 55, Issue 4
Phase transition, Raman spectra, infrared spectra, band gap and microwave dielectric properties of low temperature firing (Na0.5xBi1−0.5x)(MoxV1−x)O4 solid solution ceramics with scheelite structures
journal, January 2011
- Zhou, Di; Pang, Li-Xia; Wang, Hong
- Journal of Materials Chemistry, Vol. 21, Issue 45
Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups
journal, November 2012
- Hinton, Geoffrey; Deng, Li; Yu, Dong
- IEEE Signal Processing Magazine, Vol. 29, Issue 6
Solving the quantum many-body problem with artificial neural networks
journal, February 2017
- Carleo, Giuseppe; Troyer, Matthias
- Science, Vol. 355, Issue 6325
Transparency by Design: Closing the Gap Between Performance and Interpretability in Visual Reasoning
conference, June 2018
- Mascharka, David; Tran, Philip; Soklaski, Ryan
- 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Synthesis and characterisation of bismuth(III) vanadate
journal, June 2005
- Gotić, M.; Musić, S.; Ivanda, M.
- Journal of Molecular Structure, Vol. 744-747
Combinatorial alloying improves bismuth vanadate photoanodes via reduced monoclinic distortion
journal, January 2018
- Newhouse, P. F.; Guevarra, D.; Umehara, M.
- Energy & Environmental Science, Vol. 11, Issue 9