Hello!

I am Samar (pronounced as Summer or /ˈsʌmə(ɹ)/), a PhD candidate at the University of Pennsylvania. I work under the supervision of Prof. Alejandro Ribeiro. I spend my time at Penn exploring how to make neural networks more efficient and generalizable in solving optimization problems beyond the scope of imitation learning. My research journey has included work on algorithm unrolling with descent constraints, applied to diverse settings such as constrained optimization, bilevel optimization and federated learning. I am also interested in graph learning and currently investigating generative modeling for graph-structured signals. My ongoing research aims to extend these ideas to optimization-driven generative models. Prior to joining Penn, I was a lecturer assistant at Port Said University, Egypt, from which I had received B.Sc. and M.Sc. degrees in Electrical Engineering.

Reach out at selaraby (at) seas (dot) upenn (dot) edu.

__

News
Apr. 2025: Our new paper Generative diffusion models for resource allocation in wireless networks is out
Apr. 2025: Gave a talk about our work Robust stochastically-descending unrolled networks at ICASSP 2025
Aug. 2024: Our new paper Robust stochastically-descending unrolled networks was accepted to IEEE transactions on Signal Processing.
June 2024: Gave a Talk, Robust unrolled networks, at BIRS workshop.
Oct. 2023: Presented a poster at the 2023 Fall Fourier Talks at University of Maryland.
July 2023: Presented a poster at the institute for learning-enabled optimization at scale (TILOS) at UCSD.
May 2023: A new pre-print, Stochastic unrolled federated learning, is out.
May 2023: A short talk for Space-time graph neural networks with stochastic graph perturbations is now available.
Feb. 2023: Our paper Space-time graph neural networks with stochastic graph perturbations was accepted to ICASSP 2023.
Oct. 2022: A new pre-print is out, Space-time graph neural networks with stochastic graph perturbations.
Apr. 2022: Our paper Space-time graph neural networks was accepted to ICLR 2022. A five-minute talk is now available.