Topics for Master theses

• Understanding and implementing predictive information criteria for Bayesian models
• Asymptotic behavior of the posterior in overfitted (deep) mixture models
• Posterior concentration rates for Bayesian high-dimensional sparse additive models
• Uncertainty quantification for deep learning
• Using stacking to average distributional regression models
• Measuring explained variance in structured additive distributional regression
• Comparisons and implementation of non-local shrinkage priors
• Approximated Bayesian multivariate spatial factor analysis
• Targeted sampling based on epistemic uncertainty for improving predictive performance in DL tasks with costly data acquisition
• Uncertainty quantification for complex-valued deep neural networks
• Variational inference for distributional regression
• Informed shrinkage in Bayesian graphical modeling via external networks
• Basis selection in Bayesian semi-parametric regression using increasing shrinkage priors

Contact: Prof. Dr. Nadja Klein