REDML: Machine learning techniques for X-ray-crystallography data reduction and real-time feedback
by
OFLG/402
Abstract: In this talk, we present our latest advancements in real-time processing of X-ray diffraction images produced by the Synchrotron-light-source facility of PSI. We will delve into the techniques we have developed, addressing the challenges encountered, particularly the management of the high volume of images generated every second (kHz regime). Learn about the tools we have employed to overcome these obstacles and gain insights into their practical applications, including real-time feedback and online data reduction.
Our results show that it is possible to achieve real-time indexing and Physics-informed data reduction without sacrificing quality. On the contrary, running the standard processing pipeline after our data reduction pre-processing further improves the estimation of the electron density. These new methods opens the door to novel dynamic experiments that were elusive at this point.
These achievements are a result of our involvement in the RED-ML project, part of the 4th SDSC call.
Zoom ID: https://psich.zoom.us/j/62583239493?pwd=eXJaV0M4bmEra1IxRUwwU0hPUENBQT09
Meeting-ID: 625 8323 9493 Passcode: 026828
Laboratory for Simulation and Modeling
SDSC Hub at PSI