Mooij, Milan de (2024) Deep Learning Based Denoising of X-ray Scattering Data with Calibrated Confidence Limits. Master's Thesis / Essay, Artificial Intelligence.
|
Text
MdeMooijs3513092MScthesis.pdf Download (16MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
Toestemming.pdf Restricted to Registered users only Download (175kB) |
Abstract
Rapid advancements in experimental techniques have necessitated the development of equally swift and reliable data processing methods. This thesis addresses the challenge of denoising X-ray scattering data, an essential step for enhancing the speed and accuracy of scientific experiments. In this thesis, we implement a deep supervised learning approach trained exclusively on synthetic data. We demonstrate not only the method’s capability to generate reliable predictions on synthetic data but also its effectiveness on experimental data. This innovative approach has the potential to significantly accelerate the experimental process related to X-ray analysis, drastically reducing the time required for measurements. Additionally, we extend the denoising pipeline with a technique that provides calibrated confidence limits on point estimates, enabling researchers to quickly judge the reliability of the model’s estimations. For this extension, we deploy a quantile regression model and calibrate the estimated quantiles using Conformal Prediction. This calibration ensures that the confidence intervals are valid according to a user-specified coverage rate. The integration of machine learning into X-ray data analysis as described in this thesis promises substantial improvements in the speed and precision of scientific research. The adoption of these advanced methods can lead to a new standard in data processing, where high-throughput and highf idelity analysis become the norm. Future work will focus on expanding the applicability of these techniques to more complex experimental data and exploring real-time implementation during live experiments, potentially revolutionizing the field of experimental physics.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Sabatelli, M. and Portale, G. |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 28 Jun 2024 06:45 |
| Last Modified: | 27 Mar 2025 10:30 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/32816 |
Actions (login required)
 |
View Item |