A Fully Convolutional Approach to Denoising Structural Dynamics Data from X-Ray Photon Correlation Spectroscopy

A new fully convolutional denoising autoencoder (FC-DAE) has been developed for denoising two-time intensity-intensity correlation functions in X-ray photon correlation spectroscopy (XPCS). This model uniquely accommodates inputs of arbitrary dimensions while preserving correlation structures, trained on data from NSLS-II beamlines with techniques to enhance dataset diversity and mitigate overfitting.

The introduction of the FC-DAE is significant as it enhances the ability to recover complex dynamical features in low signal-to-noise conditions, thereby improving the reliability of structural fidelity assessments in XPCS applications.