Article Effects of data bias on machine-learning–based material discovery using experimental property data

Masaya Kumagai ; Yuki Ando ; Atsumi Tanaka ; Koji Tsuda SAMURAI ORCID (National Institute for Materials ScienceROR) ; Yukari Katsura SAMURAI ORCID (National Institute for Materials ScienceROR) ; Ken Kurosaki ORCID

Effects of data bias on machine-learning based material discovery using experimental property data.pdf
Download file (6.13 MB)
Download as zip (5.63 MB)
Collection

Citation
Masaya Kumagai, Yuki Ando, Atsumi Tanaka, Koji Tsuda, Yukari Katsura, Ken Kurosaki. Effects of data bias on machine-learning–based material discovery using experimental property data. Science and Technology of Advanced Materials: Methods. 2022, 2 (1), 302-309. https://doi.org/10.1080/27660400.2022.2109447
SAMURAI

Description:

(abstract)

本研究では、機械学習（ML）を用いた材料インフォマティクス（MI）において、大規模な材料データの偏りが機械学習モデルに与える影響を明らかにした。特に、適用領域の概念とクラスタリングを組み合わせ、データの偏りがMLモデルの予測誤差と信頼性に影響を与えることを示した。適用領域内での予測は信頼性が高い一方、適用領域外では予測の信頼性が低下する。この結果から、MLモデルが信頼できる材料探索の範囲は限られていることがわかったが、それでも新材料の提案は可能である。

Rights:

Keyword: Machine learning, material informatics, large-scale material data, data bias

Date published: 2022-12-31

Publisher: Informa UK Limited

Journal:

  • Science and Technology of Advanced Materials: Methods (ISSN: 27660400) vol. 2 issue. 1 p. 302-309

Funding:

  • JSPS KAKENHI JP20K22466

Manuscript type: Publisher's version (Version of record)

MDR DOI:

First published URL: https://doi.org/10.1080/27660400.2022.2109447

Related item:

Other identifier(s):

Contact agent:

Updated at: 2024-11-16 08:31:01 +0900

Published on MDR: 2024-11-16 08:31:01 +0900

Filename Size
Filename Effects of data bias on machine-learning based material discovery using experimental property data.pdf (Thumbnail)
application/pdf 		Size 6.13 MB Detail