Efficient Exploration of Microstructure-Property Spaces via Active Learning
In materials design, supervised learning plays an important role for optimization and inverse modeling of microstructure-property relations. To successfully apply supervised learning models, it is essential to train them on suitable data. Here, suitable means that the data covers the microstructure...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2022-02-01
|
| Series: | Frontiers in Materials |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fmats.2021.824441/full |
| _version_ | 1819279142896533504 |
|---|---|
| author | Lukas Morand Norbert Link Tarek Iraki Johannes Dornheim Johannes Dornheim Dirk Helm |
| author_facet | Lukas Morand Norbert Link Tarek Iraki Johannes Dornheim Johannes Dornheim Dirk Helm |
| author_sort | Lukas Morand |
| collection | DOAJ |
| description | In materials design, supervised learning plays an important role for optimization and inverse modeling of microstructure-property relations. To successfully apply supervised learning models, it is essential to train them on suitable data. Here, suitable means that the data covers the microstructure and property space sufficiently and, especially for optimization and inverse modeling, that the property space is explored broadly. For virtual materials design, typically data is generated by numerical simulations, which implies that data pairs can be sampled on demand at arbitrary locations in microstructure space. However, exploring the space of properties remains challenging. To tackle this problem, interactive learning techniques known as active learning can be applied. The present work is the first that investigates the applicability of the active learning strategy query-by-committee for an efficient property space exploration. Furthermore, an extension to active learning strategies is described, which prevents from exploring regions with properties out of scope (i.e., properties that are physically not meaningful or not reachable by manufacturing processes). |
| first_indexed | 2024-12-24T00:23:12Z |
| format | Article |
| id | doaj.art-c76af82cabb44e9cbf567d3e08c18244 |
| institution | Directory Open Access Journal |
| issn | 2296-8016 |
| language | English |
| last_indexed | 2024-12-24T00:23:12Z |
| publishDate | 2022-02-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Materials |
| spelling | doaj.art-c76af82cabb44e9cbf567d3e08c182442022-12-21T17:24:32ZengFrontiers Media S.A.Frontiers in Materials2296-80162022-02-01810.3389/fmats.2021.824441824441Efficient Exploration of Microstructure-Property Spaces via Active LearningLukas Morand0Norbert Link1Tarek Iraki2Johannes Dornheim3Johannes Dornheim4Dirk Helm5Fraunhofer Institute for Mechanics of Materials IWM, Freiburg, GermanyIntelligent Systems Research Group ISRG, Karlsruhe University of Applied Sciences, Karlsruhe, GermanyIntelligent Systems Research Group ISRG, Karlsruhe University of Applied Sciences, Karlsruhe, GermanyIntelligent Systems Research Group ISRG, Karlsruhe University of Applied Sciences, Karlsruhe, GermanyInstitute for Applied Materials—Computational Materials Sciences IAM-CMS, Karlsruhe Institute of Technology, Karlsruhe, GermanyFraunhofer Institute for Mechanics of Materials IWM, Freiburg, GermanyIn materials design, supervised learning plays an important role for optimization and inverse modeling of microstructure-property relations. To successfully apply supervised learning models, it is essential to train them on suitable data. Here, suitable means that the data covers the microstructure and property space sufficiently and, especially for optimization and inverse modeling, that the property space is explored broadly. For virtual materials design, typically data is generated by numerical simulations, which implies that data pairs can be sampled on demand at arbitrary locations in microstructure space. However, exploring the space of properties remains challenging. To tackle this problem, interactive learning techniques known as active learning can be applied. The present work is the first that investigates the applicability of the active learning strategy query-by-committee for an efficient property space exploration. Furthermore, an extension to active learning strategies is described, which prevents from exploring regions with properties out of scope (i.e., properties that are physically not meaningful or not reachable by manufacturing processes).https://www.frontiersin.org/articles/10.3389/fmats.2021.824441/fullactive learningadaptive samplingdata generationinverse modelingmaterials designmembership query synthesis |
| spellingShingle | Lukas Morand Norbert Link Tarek Iraki Johannes Dornheim Johannes Dornheim Dirk Helm Efficient Exploration of Microstructure-Property Spaces via Active Learning Frontiers in Materials active learning adaptive sampling data generation inverse modeling materials design membership query synthesis |
| title | Efficient Exploration of Microstructure-Property Spaces via Active Learning |
| title_full | Efficient Exploration of Microstructure-Property Spaces via Active Learning |
| title_fullStr | Efficient Exploration of Microstructure-Property Spaces via Active Learning |
| title_full_unstemmed | Efficient Exploration of Microstructure-Property Spaces via Active Learning |
| title_short | Efficient Exploration of Microstructure-Property Spaces via Active Learning |
| title_sort | efficient exploration of microstructure property spaces via active learning |
| topic | active learning adaptive sampling data generation inverse modeling materials design membership query synthesis |
| url | https://www.frontiersin.org/articles/10.3389/fmats.2021.824441/full |
| work_keys_str_mv | AT lukasmorand efficientexplorationofmicrostructurepropertyspacesviaactivelearning AT norbertlink efficientexplorationofmicrostructurepropertyspacesviaactivelearning AT tarekiraki efficientexplorationofmicrostructurepropertyspacesviaactivelearning AT johannesdornheim efficientexplorationofmicrostructurepropertyspacesviaactivelearning AT johannesdornheim efficientexplorationofmicrostructurepropertyspacesviaactivelearning AT dirkhelm efficientexplorationofmicrostructurepropertyspacesviaactivelearning |