| About this Abstract |
| Meeting |
TMS Specialty Congress 2025
|
| Symposium
|
Joint Sessions of AIM, ICME, & 3DMS
|
| Presentation Title |
X-Ray Diffraction Analysis Using TensorFlow and FAIR Data Pipelines |
| Author(s) |
Finley Holt, Daniel Savage, Mohommad Mehdi, Weiqi Yue, Pawan Tripathi, Matthew Willard, Frank Ernst, Roger French |
| On-Site Speaker (Planned) |
Daniel Savage |
| Abstract Scope |
X-ray diffraction generates vast, complex datasets of material behavior that demand scalable and flexible scientific analyses. The efficient management and manipulation of image and histogram data has largely been addressed in the development of the TensorFlow package for ML and AI. In this talk we will explore using our newly developed FAIRshake package, an end-to-end, modular framework that interfaces with FAIRified diffraction data using the TensorFlow Dataset API, to perform data manipulation and analysis. TensorFlow allows data transformation (e.g dark corrections, azimuthal integration, analysis) to be performed using standard TensorFlow dataset tools. The autotuning capability of TensorFlow enables excellent performance, from desktops to HPC environments, through dynamic dataset streaming and parallelization. TensorFlow datasets are shown through examples to be especially attractive for scientific analysis that can natively utilize tensor representations of data; bringing into focus the question: “How should scientific codes be interacting with FAIR data?” |
| Proceedings Inclusion? |
Definite: Post-meeting proceedings |