Sergio Rozada

Assistant Professor at ECE, URJC. Interested in dynamical systems, reinforcement learning, and optimization.

I am an Assistant Professor in the Department of Signal Theory and Communications at Universidad Rey Juan Carlos (URJC) in Madrid, where I teach signal processing, optimization, and machine learning.

I earned my PhD at URJC in 2025, having the honor to be advised by Prof. Antonio G. Marques. My thesis, Low-Rank Methods in Reinforcement Learning, received the Young Researchers and Innovators Award for the best engineering dissertation at URJC. During my doctorate I was a visiting researcher in Prof. Alejandro Ribeiro’s Alelab at the University of Pennsylvania. I received my MSc in Data Science from City, University of London in 2018, and my BSc in Electronics Engineering and Industrial Automation from the University of Oviedo in 2015. Before returning to academia, I spent some years in industry at ArcelorMittal, BBVA, Seedtag, and Meta, developing data and machine-learning products.

My research interests encompass reinforcement learning, generative diffusion models, graph signal processing, and tensor and low-rank methods. I exploit the low-rank structure hidden in value functions and policies to make reinforcement learning tractable in large state-action spaces, and I recast dynamic programming and generative diffusion through the lens of graph signal processing. These perspectives turn classical signal processing tools into parsimonious models with provable guarantees for sequential decision making and for data defined over irregular domains.

My ultimate goal is to use insights from signal processing and optimization to characterize the mathematical foundations, develop novel architectures, and devise efficient learning procedures that let us understand and control complex dynamical systems.

News

  • 2026-02-04 Thesis award — My PhD thesis, Low-Rank Methods in Reinforcement Learning, received the Young Researchers and Innovators Award for the Best Engineering Dissertation at URJC in 2025.
  • 2026-01-17 Papers accepted at ICASSP 2026 — Two papers on graph signal diffusion have been accepted at ICASSP 2026, focusing on generative modeling and diffusion processes over graph-structured data.
  • 2025-09-22 Paper accepted at NPGML workshop, Neurips — We are presenting our GSP perspective on RL in the NPGML workshop at Neurips! See you in San Diego.
  • 2025-09-16 Paper accepted at CAMSAP — Our paper on a GSP perspective on RL has been accepted at CAMSAP.