Research

My field of research is geometric optimization in imaging with special interest in statistical manifolds, information geometry and everything related to Bregman-divergences. Currently, I am working on multilevel-methods to efficiently accelerate mirror descent on high-dimensional Riemannian manifolds.

Publications

  • Elshiaty, Yara and Petra, Stefania Multilevel Bregman Proximal Gradient Descent, SIAM Journal on Imaging Sciences, 2026. [DOI] [arXiv]
  • Vanmaele, F. and Elshiaty, Y. and Petra, S. Multilevel Optimization: Geometric Coarse Models and Convergence Analysis, International Conference on Scale Space and Variational Methods in Computer Vision (SSVM), 2025. [DOI] [arXiv]
  • Elshiaty, Y. and Vanmaele, F. and Petra, S. Information Geometry of Exponentiated Gradient: Convergence beyond L-Smoothness, International Conference on Scale Space and Variational Methods in Computer Vision (SSVM), 2025. [DOI] [arXiv]

Contact

Email elshiaty [at] math [dot] uni-heidelberg [dot] de
Office Mathematikon, Office 4/229
Phone +49 6221 5414 839
ORCID 0000-0001-8705-2484
Social Profile scholar.google.com/citations?user=PtUCl8IAAAAJ&hl=en