Advances in Markov Chain Monte Carlo for Bayesian Deep Learning for Uncertainty Quantification at Scale

MCML Authors

Emanuel Sommer

Dr.

→ Group David Rügamer
 Statistics, Data Science and Machine Learning

Abstract

This dissertation advances scalable sampling-based Bayesian inference for deep neural networks, enabling more accurate and efficient uncertainty quantification. It introduces novel sampling methods that overcome computational and theoretical challenges, extending Bayesian inference to large architectures such as Vision Transformers. By improving sampling efficiency, robustness, and automation, the work establishes practical alternatives to traditional optimization-based approximations. (Shortened.)

phdthesis Som26

---

Dissertation

LMU München. Sep. 2026

---

Authors

E. Sommer

---

Links

DOI

---

Research Area

 A1 | Statistical Foundations & Explainability