Data-driven microscopy analysis
Unsupervised learning for identifying meaningful spectral and structural variation in multidimensional electron-microscopy data.
This research applies dimensionality reduction and unsupervised clustering to microscopy datasets where meaningful spectral features are distributed across many pixels or dimensions. The approach supports exploratory analysis without prior labels and connects the resulting clusters to their real-space distributions.
The associated open workflow is available through DRCA.
Publications
- J. Ryu, et al., Dimensionality reduction and unsupervised clustering for EELS-SI, Ultramicroscopy (2021).
- J. Ryu, et al., Correlative study between the local atomic and electronic structures of amorphous carbon materials via 4D-STEM and STEM-EELS, Applied Physics Letters (2022).
- S. Lee, J. Ryu, et al., In situ transmission electron microscopy visualization of electric-field-induced phase transitions at the morphotropic phase boundary in Hf0.5Zr0.5O2, ACS Nano (2026).