Dr. Ankur Singha
Dr. Ankur Singha
Postdoctoral Researcher
Affiliation: BIFOLD
Ankur Singha is a Postdoctoral Research Associate in the Machine Learning Group at Technische Universität Berlin. His research focuses on developing advanced generative AI models to improve sampling methods in lattice field theory, with the broader goal of enhancing the efficiency and scalability of simulations for lattice QCD.
He received his doctoral degree in Physics from IIT Kanpur, India, in 2023. During his PhD, he worked on conditional generative adversarial networks and conditional flow-based models for simulating lattice systems, including the Gross–Neveu model, Phi-4 theory, and two-dimensional U(1) gauge theory. His current research is focused on the development of multilevel generative samplers for lattice field theory.
My research interests center on the development of machine learning methods for lattice-based physical systems. I am particularly interested in generative modeling for efficient sampling, multiscale and multilevel approaches inspired by renormalization-group ideas, and reinforcement-learning methods for optimization problems in gauge theories. More broadly, I aim to explore how modern AI techniques can help address computational challenges in statistical physics, lattice field theory, and lattice gauge theory.
- Lattice field theory
- Lattice gauge theory
- Statistical physics
- generative models
- Normalizing flows
- Multilevel methods
- Reinforcement learning