Compositionally restricted attention-based network for materials property predictions

Abstract In this paper, we demonstrate an application of the Transformer self-attention mechanism in the context of materials science. Our network, the Compositionally Restricted Attention-Based network (CrabNet), explores the area of structure-agnostic materials property predictions when only a che...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Anthony Yu-Tung Wang, Steven K. Kauwe, Ryan J. Murdock, Taylor D. Sparks
Formato:	article
Lenguaje:	EN
Publicado:	Nature Portfolio 2021
Materias:	Materials of engineering and construction. Mechanics of materials TA401-492 Computer software QA76.75-76.765
Acceso en línea:	https://doaj.org/article/84ef20134dd54902b0ccba0b112204ea
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:84ef20134dd54902b0ccba0b112204ea
record_format	dspace
spelling	oai:doaj.org-article:84ef20134dd54902b0ccba0b112204ea2021-12-02T15:00:50ZCompositionally restricted attention-based network for materials property predictions10.1038/s41524-021-00545-12057-3960https://doaj.org/article/84ef20134dd54902b0ccba0b112204ea2021-05-01T00:00:00Zhttps://doi.org/10.1038/s41524-021-00545-1https://doaj.org/toc/2057-3960Abstract In this paper, we demonstrate an application of the Transformer self-attention mechanism in the context of materials science. Our network, the Compositionally Restricted Attention-Based network (CrabNet), explores the area of structure-agnostic materials property predictions when only a chemical formula is provided. Our results show that CrabNet’s performance matches or exceeds current best-practice methods on nearly all of 28 total benchmark datasets. We also demonstrate how CrabNet’s architecture lends itself towards model interpretability by showing different visualization approaches that are made possible by its design. We feel confident that CrabNet and its attention-based framework will be of keen interest to future materials informatics researchers.Anthony Yu-Tung WangSteven K. KauweRyan J. MurdockTaylor D. SparksNature PortfolioarticleMaterials of engineering and construction. Mechanics of materialsTA401-492Computer softwareQA76.75-76.765ENnpj Computational Materials, Vol 7, Iss 1, Pp 1-10 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Materials of engineering and construction. Mechanics of materials TA401-492 Computer software QA76.75-76.765
spellingShingle	Materials of engineering and construction. Mechanics of materials TA401-492 Computer software QA76.75-76.765 Anthony Yu-Tung Wang Steven K. Kauwe Ryan J. Murdock Taylor D. Sparks Compositionally restricted attention-based network for materials property predictions
description	Abstract In this paper, we demonstrate an application of the Transformer self-attention mechanism in the context of materials science. Our network, the Compositionally Restricted Attention-Based network (CrabNet), explores the area of structure-agnostic materials property predictions when only a chemical formula is provided. Our results show that CrabNet’s performance matches or exceeds current best-practice methods on nearly all of 28 total benchmark datasets. We also demonstrate how CrabNet’s architecture lends itself towards model interpretability by showing different visualization approaches that are made possible by its design. We feel confident that CrabNet and its attention-based framework will be of keen interest to future materials informatics researchers.
format	article
author	Anthony Yu-Tung Wang Steven K. Kauwe Ryan J. Murdock Taylor D. Sparks
author_facet	Anthony Yu-Tung Wang Steven K. Kauwe Ryan J. Murdock Taylor D. Sparks
author_sort	Anthony Yu-Tung Wang
title	Compositionally restricted attention-based network for materials property predictions
title_short	Compositionally restricted attention-based network for materials property predictions
title_full	Compositionally restricted attention-based network for materials property predictions
title_fullStr	Compositionally restricted attention-based network for materials property predictions
title_full_unstemmed	Compositionally restricted attention-based network for materials property predictions
title_sort	compositionally restricted attention-based network for materials property predictions
publisher	Nature Portfolio
publishDate	2021
url	https://doaj.org/article/84ef20134dd54902b0ccba0b112204ea
work_keys_str_mv	AT anthonyyutungwang compositionallyrestrictedattentionbasednetworkformaterialspropertypredictions AT stevenkkauwe compositionallyrestrictedattentionbasednetworkformaterialspropertypredictions AT ryanjmurdock compositionallyrestrictedattentionbasednetworkformaterialspropertypredictions AT taylordsparks compositionallyrestrictedattentionbasednetworkformaterialspropertypredictions
_version_	1718389149405282304

Compositionally restricted attention-based network for materials property predictions

Ejemplares similares