- Abstract
- Zusammenfassung
- Declaration and Acknowledgements
- Symbols and Abbreviations
- Contents
- 1. Introduction
- 2. Tomography problem and Singular-Value Decomposition (SVD) analysis
- 3. Theory
- 3.1. Bayesian probability theory
- 3.1.1. Introduction
- 3.1.2. Marginalization of conditional probability distribution
- 3.1.3. Multivariate Normal (MVN) distribution
- 3.1.4. Likelihood function
- 3.1.5. Prior and Regularization
- 3.1.6. Non-negative constraint realized by truncated MVN
- 3.1.7. Optimization of model assumption: Bayesian Occam’s razor
- 3.1.8. Maximum Entropic regularization
- 3.2. Gaussian Process (GP)
- 3.3. Non-stationary Gaussian Process
- 3.4. Non-stationary Gaussian Process Tomography (NSGPT)
- 3.1. Bayesian probability theory
- 4. Non-stationary GP tomography (NSGPT) of soft X-ray diagnostics
- 4.1. Selection criterion of reconstruction region
- 4.2. Benchmark by simulation
- 4.3. Application to the soft X-ray diagnostics in W7-AS stellarator
- 4.4. SVD analysis for identification of mode structures
- 4.5. Application to the soft X-ray diagnostics in TJ-II stellarator
- 4.6. Discussion and conclusion on SXR reconstructions
- 5. Implementation of Gaussian Process tomography to bolometer diagnostics
- 6. Conclusion and Summary
- 7. Outlook
- 8. Appendices
- 9. Bibliography
