EELSfitter</div>
Machine Learning for High-Dimensional Microscopy Data
We develop machine-learning methods to analyse high-dimensional electron microscopy datasets, with a focus on robust and uncertainty-aware interpretation of EELS measurements.
Our work focuses on electron energy-loss spectroscopy and related multidimensional microscopy data, where weak spectral features must be separated from background, noise, and instrumental artefacts.
</section>
Context
Modern electron microscopy produces rich multidimensional datasets, including EELS spectrum images, spatially resolved spectroscopic maps, and 4D-STEM measurements. These data contain information about electronic structure, excitations, dielectric response, chemical bonding, and nanoscale heterogeneity.
Challenge and motivation
In EELS, weak spectral features are often hidden by strong backgrounds, such as the zero-loss peak in the low-loss regime or the pre-edge background in core-loss spectroscopy. Standard fitting approaches can be sensitive to user choices, especially when applied to large spectrum images.
EELSfitter
EELSfitter is our open-source Python framework for robust analysis and interpretation of electron energy-loss spectroscopy data. It adapts machine-learning strategies developed in high-energy physics, including feed-forward neural networks for unbiased regression in multidimensional problems.