Versatile automated domain mapping of 4D-STEM data utilizing ML algorithms and Bayesian statistics

DOI: 10.1017/S1431927622011308

Authors: A. Bridger (Diamond Light Source; Science and Technology Facilities Council; University of Oxford) , M. Danaie (Diamond Light Source) , K. Butler (Science and Technology Facilities Council) , T. Wood (Science and Technology Facilities Council) , W. I. F. David (Science and Technology Facilities Council; University of Oxford)
Co-authored by industrial partner: No

Type: Conference Paper
Conference: Microscopy & Microanalysis 2022: C04 - Artificial Intelligence, Instrument Automation, and High-dimensional Data Analytics for Microscopy and Microanalysis
Peer Reviewed: No

State: Published (Approved)
Published: August 2022

Abstract: This work outlines the development of a workflow that combines both informed and data-driven approaches to process and classify 4D-STEM data utilising signal dimensionality reduction obtained from unsupervised machine learning algorithms. Using these ‘latent space’ representations of the scanned region, we can then apply clustering algorithms to identify regions of similar character. By combining the results of multiple autonomous computations, we would be able to map domains within the crystal and associate a level of confidence to given classifications. This confidence is then used to iteratively improve the learning algorithms and the performance of their dimensionality reduction. Incorporating this as a pre-processing unsupervised workflow would drastically improve the ability to characterise the nano-scale structure of materials, both through production of significantly signal-boosted diffraction data and reduction in laborious manual investigation.

Subject Areas: Information and Communication Technology, Materials


Technical Areas:

Added On: 27/07/2022 09:43

Discipline Tags:

Information & Communication Technologies Data processing Mathematics

Technical Tags: