EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Dataset-Dependent Performance of Active Learning Acquisition Functions for Multi-Objective Materials Optimization

Jingjing Yan, Qing Zhao, Huiling He, Zhiqiang Mu, Haoyu Song, Yiming Chen, Xiaobing Hu, Jincheng Wang

Materials · 07.11.2026 · 19 (21) · 4736

DOI: 10.3390/ma19214736 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Materials, online erschienen am 07.11.2026. Verfasst von Jingjing Yan und 7 weiteren Autor:innen. Beteiligte Einrichtung: School of Architecture, Kashi University, Kashi 844000, China. Die Arbeit nennt 43 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Kashi University Cultivation Project, Open Research Fund of Shaanxi Laboratory of Advanced Materials, the Natural Science Basic Research Program-General Projects-Youth Projects of Shaanxi Province (Category C).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Active learning (AL) promises accelerated multi-objective materials discovery, yet acquisition-function choice is often made without matching dataset, budget, and target metric. We retrospectively compared EI-random, EI-cent, EI-maxmin, and EI-prod on Al alloy, steel, and Cu alloy datasets across 100 randomized cold-start trials, using hypervolume, true Pareto-sample discovery, inverted generational distance, balanced-performance score, and iterations to 95% target. EI-prod is best overall on Al dataset; EI-maxmin improves Pareto-front recovery and convergence on steel and Cu datasets; EI-cent gives limited, inconsistent gains. No strategy dominates all datasets or budgets. These results provide a metric-aware, outcome-specific selection guide: use EI-prod for early balanced candidates, and EI-maxmin for broader Pareto-front exploration. They also warn that similar balanced performance does not ensure equivalent front reconstruction and that dataset-specific rankings should inform, not dictate, AL design in materials screening. This reframes acquisition selection as a decision-making, not universal-ranking, problem.

Schlagwörter

Stichwörter der Autor:innen

acquisition functionsactive learningmulti-objective optimization

Autor:innen und Institutionen

  • School of Architecture, Kashi University, Kashi 844000, China.
  • School of Civil Engineering, Kashi University, Kashi 844000, China.
  • Xi'an Rare Metal Materials Institute Co., Ltd., Xi'an 710016, China.
  • State Key Laboratory of Solidification Processing, Northwestern Polytechnical University, Xi'an 710072, China.
  • School of Metallurgical Engineering, National and Local Joint Engineering Research Center for Functional Materials Processing, Xi'an University of Architecture and Technology, Xi'an 710055, China.

Förderung

  • Kashi University Cultivation Project
  • Open Research Fund of Shaanxi Laboratory of Advanced Materials
  • the Natural Science Basic Research Program-General Projects-Youth Projects of Shaanxi Province (Category C)
  • Xi'an University of Architecture and Technology
  • Natural Science Basic Research Program-General Projects-Youth Projects of Shaanxi Province
  • Xi’an University of Architecture and Technology

Zitieren

APA

Yan, J., Zhao, Q., He, H., Mu, Z., Song, H., Chen, Y., Hu, X., & Wang, J. (2026). Dataset-Dependent Performance of Active Learning Acquisition Functions for Multi-Objective Materials Optimization. Materials, 19(21), 4736. https://doi.org/10.3390/ma19214736

Vancouver

Yan J, Zhao Q, He H, Mu Z, Song H, Chen Y, et al. Dataset-Dependent Performance of Active Learning Acquisition Functions for Multi-Objective Materials Optimization. Materials. 2026;19(21):4736. doi:10.3390/ma19214736

BibTeX
@article{Yan2026,
  author = {Yan, Jingjing and Zhao, Qing and He, Huiling and Mu, Zhiqiang and Song, Haoyu and Chen, Yiming and Hu, Xiaobing and Wang, Jincheng},
  title = {Dataset-Dependent Performance of Active Learning Acquisition Functions for Multi-Objective Materials Optimization},
  journal = {Materials},
  year = {2026},
  volume = {19},
  number = {21},
  pages = {4736},
  doi = {10.3390/ma19214736}
}
Kostenlos starten →