en
Kim Andrea Nicoli
© F. Noak/BIFOLD
Dr. Kim Andrea Nicoli

Website

Affiliations
  • Helmholtz Institute for Radiation and Nuclear Physics
  • TRA Matter
Research topics
  • Machine Learning
  • Variational Inference
  • Lattice Field Theory
  • Quantum Computing
My research lies at the intersection of physics and modern machine learning, with a focus on generative models such as normalizing flows, diffusion, and autoregressive models, and their application to physics. I am also interested in theoretical aspects of machine learning, including geometric aspects of deep learning, density estimation, optimal transport, sampling tasks, and modeling of multimodal distributions. Additionally, I explore Bayesian inference techniques, such as Gaussian processes and Bayesian optimization, for solving physics-related problems. These techniques have potential in lattice quantum field theory, soft-condensed matter, quantum chemistry, molecular science, and modern quantum computing.
Selected publications

Nicoli, K. A., Nakajima, S., Strodthoff, N., Samek, W., Müller, K. R., & Kessel, P. (2020). Asymptotically unbiased estimation of physical observables with neural samplers. Physical Review E, 101(2), 023304.

Nicoli, K. A., Anders, C. J., Funcke, L., Hartung, T., Jansen, K., Kessel, P., ... & Stornati, P. (2021). Estimation of thermodynamic observables in lattice field theories with deep generative models. Physical review letters, 126(3), 032001.

Vaitl, L., Nicoli, K. A., Nakajima, S., & Kessel, P. (2022). Gradients should stay on path: better estimators of the reverse-and forward KL divergence for normalizing flows. Machine Learning: Science and Technology, 3(4), 045006.

Vaitl, L., Nicoli, K. A., Nakajima, S., & Kessel, P. (2022, June). Path-gradient estimators for continuous normalizing flows. In International Conference on Machine Learning (pp. 21945-21959). PMLR

 
 

 

 
 

 

Kim Andrea Nicoli
© F. Noak/BIFOLD
Dr. Kim Andrea Nicoli

Website

-

Room 2.013

Nussallee 14-16

53111 Bonn