EUVIMEDEUVIMED entdecken
Crossref

Reward driven discovery of the optimal microstructure representations with invariant variational autoencoders

Boris Slautin, Kamyar Barakati iD, Hiroshi Funakubo, Maxim Ziatdinov, Vladimir Shvartsman iD, Doru Lupascu, Sergei V Kalinin

Nanotechnology · 17.11.2026

DOI: 10.1088/1361-6528/aec2fd ↗

Kurz gefasst

Fachartikel in der Zeitschrift Nanotechnology, online erschienen am 17.11.2026. Verfasst von Boris Slautin und 6 weiteren Autor:innen. Beteiligte Einrichtung: Department of Materials Science and Engineering, The University of Tennessee Knoxville, 2641 Osprey Vista Way, Knoxville, Tennessee, 37996-4519, Unit…. Gefördert von: NSF MRSEC, U.S. Department of Energy, Pacific Northwest National Laboratory.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Extracting physically meaningful information from microscopy images requires selecting suitable structural descriptors and constructing low-dimensional representations. However, the quality of these representations strongly depends on the choice of descriptors, model architecture, and hyperparameters, which are typically selected empirically. Variational Autoencoders (VAEs) provide a powerful means of constructing such low-dimensional representations, but their performance heavily depends on multiple non-myopic design choices, which are often optimized through trial-and-error and empirical analysis. To enable automated and unbiased optimization of VAE workflows, we investigated reward-based strategies for evaluating latent space representations. Using Piezoresponse Force Microscopy imaging as a model system, we examined multiple reward functions for automated VAE workflow optimization. Our analysis shows that Gaussian Mixture Models (GMM) and Bayesian Gaussian Mixture Models (BGMM) provide a basis for constructing reward functions that balance representation quality, complexity, and cluster separability. In the demonstrated case, the ICL-based reward identified four dominant components corresponding to distinct microstructural configurations, illustrating how reward-based optimization can guide the discovery of parsimonious representations.

Schlagwörter

Stichwörter der Autor:innen

microscopy image analysispiezoresponse force microscopyrepresentation learningreward-driven learningvariational autoencoders

Autor:innen und Institutionen

  • Department of Materials Science and Engineering, The University of Tennessee Knoxville, 2641 Osprey Vista Way, Knoxville, Tennessee, 37996-4519, United States.
  • University of Tennessee Knoxville College of Engineering, 2641 Osprey Vista Way, Knoxville, Tennessee, 37996, United States.
  • Tokyo Institute of Technology, 2 Chome-12-1 Ookayama, Meguro, Tokyo, 152-8550, Japan.
  • Pacific Northwest National Laboratory, 902 Battelle Boulevard, Richland, Richland, Washington, 99352, United States.
  • Ingenieurwissenschaften Institut fuer Materialwissenschaft, Universität Duisburg-Essen Fakultät für Ingenieurwissenschaften, Universitaetstrasse 15, 45141 Essen, Duisburg, 47048, Germany.
  • Institute for Materials Science, University of Duisburg-Essen, Universitaetsstr. 15, Duisburg, NRW, 47057, Germany.

ORCID-Profile: Kamyar Barakati, Vladimir Shvartsman

Förderung

  • NSF MRSEC
  • U.S. Department of Energy
  • Pacific Northwest National Laboratory
  • Center for Advanced Materials and Manufacturing
  • Japan Science and Technology Agency

Zitieren

APA

Slautin, B., Barakati, K., Funakubo, H., Ziatdinov, M., Shvartsman, V., Lupascu, D., & Kalinin, S. V. (2026). Reward driven discovery of the optimal microstructure representations with invariant variational autoencoders. Nanotechnology. https://doi.org/10.1088/1361-6528/aec2fd

Vancouver

Slautin B, Barakati K, Funakubo H, Ziatdinov M, Shvartsman V, Lupascu D, et al. Reward driven discovery of the optimal microstructure representations with invariant variational autoencoders. Nanotechnology. 2026. doi:10.1088/1361-6528/aec2fd

BibTeX
@article{Slautin2026,
  author = {Slautin, Boris and Barakati, Kamyar and Funakubo, Hiroshi and Ziatdinov, Maxim and Shvartsman, Vladimir and Lupascu, Doru and Kalinin, Sergei V},
  title = {Reward driven discovery of the optimal microstructure representations with invariant variational autoencoders},
  journal = {Nanotechnology},
  year = {2026},
  doi = {10.1088/1361-6528/aec2fd}
}
Kostenlos starten →