Experimental data analysis in plasma physics

Course guarantor:

Lecturer:

Tutor:

Supervisor:

Department of Physics

Synopsis:

The goal of the course is to provide students with the opportunity to gain practical experience by solving projects in the field of data science. Several tasks focused on analyzing data from fusion experiments with magnetic plasma confinement, using various diagnostic systems (microwaves, visible spectroscopy, infrared cameras, electrical probes, etc.), give students the chance to learn how to apply Bayesian approaches, neural networks, and computations on graphics cards to obtain the required information about the plasma state. Furthermore, it introduces the advantages of applying forward and backward models in plasma diagnostics. This approach mimics workflows common in research and development projects, where the requirement is to design a method for obtaining a certain type of information from measured data.

Requirements:

Development of accounting projects - application of data analysis

Syllabus of lectures:

Syllabus of tutorials:

1. The role and types of data in fusion research, description of a physical experiment, statistical models, frequentist vs Bayesian approaches

2. The significance of forward and backward modelling of diagnostics for physics experiments

3. Bayesian models and solution methods

4. Forward modelling of optical plasma diagnostics

5. Example of an inverse problem: tomographic reconstruction - classical vs Bayesian approach

6. Integrated data analysis: merging information from multiple diagnostics

7. Gaussian processes and Bayesian optimization of black box models

8. Application of machine learning, introduction to types and use cases of neural networks

9. Convolutional neural networks for image information processing

10. Time series processing: spectral analysis, autoregressive models

11. Generative modelling of experimental data, outlier detection

12. Acceleration of computations on GPUs

Study Objective:

Acquired skills: Application of statistical data analysis using the Bayesian approach, data analysis using neural networks, acceleration of common computations using graphics cards, creation of forward and backward models, experience in participating in data science projects, and familiarity with common research and development approaches.

Study materials:

Required literature: -

Recommended literature:

[1] J. S. Bendat, A. G. Piersol, Random Data, Wiley, 2010

[2] Bishop, C. Pattern Recognition and Machine Learning, Springer, New York, 2007.

[3] I. Goodfellow, Y. Bengio, A. Courville, Deep Learning, The MIT Press, 2016

[4] Bayesian Methods for Hackers. https://dataorigami.net/Probabilistic-Programming-and-Bayesian-Methods-for-Hackers/.

[5] Statistical Rethinking 2023 - YouTube. https://www.youtube.com/playlist?list=PLDcUM9US4XdPz-KxHM4XHt7uUVGWWVSus.

[6] Welcome Bayesian Modeling and Computation in Python. https://bayesiancomputationbook.com/welcome.html.

Note:

Further information:

No time-table has been prepared for this course

The course is a part of the following study plans:

• Fyzika plazmatu a termojaderné fúze (elective course)