Zia Ur Rehman

PhD researcher in Computational Mathematics
Technical University of Munich (TUM)

About

I am a PhD researcher in Computational Mathematics at the Technical University of Munich (TUM), supervised by Prof. Dr. Gero Friesecke. My research combines optimal transport, deep learning and molecular dynamics to develop efficient methods for sampling high-dimensional Boltzmann distributions. I focus on combining mathematical theory with machine learning to develop reliable and computationally efficient methods for molecular simulation.

Research interests

Optimal transport and computational mathematics
Normalizing flows and generative modeling
Molecular dynamics and Boltzmann sampling
High-dimensional probability and machine learning

Education

2025 – present

PhD in Computational Mathematics, Technical University of Munich, Germany
Advisor: Prof. Dr. Gero Friesecke

2024

Double MSc in Mathematics for Real-world Applications (Erasmus Mundus), University of L’Aquila, Italy & Silesian University of Technology, Poland
Thesis: Generative adversarial network model for image input with noise. Supervisor: Prof. Marcin Woźniak

2020

Bachelor of Education (B.Ed.), Allama Iqbal Open University, Islamabad, Pakistan

2017

Bachelor’s studies in Mathematics, University of Peshawar, Pakistan

Publications

Sampling Boltzmann distributions via normalizing flow approximation of transport maps

Z. Ur Rehman and G. Friesecke

Transactions on Machine Learning Research (TMLR), 2026

PDFCode

Enhancing image processing with GANs in noisy environments

Z. Ur Rehman

Proc. 29th International Conference Information Society and University Studies (IVUS), Kaunas, Lithuania, 2024, pp. 310–321

PDF

Current work

A transport consistency penalty (TCP) for training Boltzmann generators, with guarantees in total variation, Wasserstein distance and effective sample size, and no need for the partition function.

Talks

Mar 2026

Sampling Boltzmann distributions via normalizing flow approximation of transport maps. 96th Annual Meeting of the International Association of Applied Mathematics and Mechanics (GAMM 2026), Stuttgart, Germany.

Jan 2025

Generative adversarial network model for image input with noise. Research Seminar, Technical University of Munich, Germany.

Contact

Department of Mathematics, School of CIT
Technical University of Munich
Boltzmannstr. 3, 85748 Garching, Germany
ziaur.rehman@tum.de