Validation of non-negative matrix factorization for assessment of atomic pair-distribution function (PDF) data in a real-time streaming context

TL;DR

This paper validates the use of matrix factorization techniques, specifically PCA and NMF, for real-time analysis of atomic pair distribution function data, demonstrating their effectiveness in identifying relevant components during in situ experiments.

Contribution

It introduces a new software infrastructure for streaming PDF data analysis and compares PCA and NMF methods for component identification in real-time.

Findings

NMF effectively identifies relevant components in streaming PDF data.

The software infrastructure enables real-time data analysis during experiments.

PCA and NMF provide complementary insights into atomic structure data.

Abstract

We validate the use of matrix factorization for the automatic identification of relevant components from atomic pair distribution function (PDF) data. We also present a newly developed software infrastructure for analyzing the PDF data arriving in streaming manner. We then apply two matrix factorization techniques, Principal Component Analysis (PCA) and Non-negative Matrix Factorization (NMF), to study simulated and experiment datasets in the context of in situ experiment.