RPI Alumnus Hongkang Li, who earned his Ph.D. in the Department of Electrical, Computer, and Systems Engineering at Rensselaer Polytechnic Institute in 2024, has been awarded the 2025 IEEE Signal Processing Society Best Ph.D. Dissertation Award. This award recognizes outstanding doctoral dissertations that demonstrate significant scientific impact and overall quality through timely studies on research questions, strong and novel methodological foundations, and contributions that advance the understanding and practice of signal processing.
His doctoral dissertation, titled “Optimization and Generalization Analysis of Advanced Neural Networks and Learning Algorithms,” focuses on developing rigorous theoretical foundations for modern deep learning models. While many advanced deep learning models, especially those used in large language models, are developed and evaluated primarily through experiments, their underlying mechanisms and reliability are often not well understood from a theoretical perspective. Dr. Li's work aims to bridge this gap by developing mathematical frameworks to analyze how advanced neural network models are optimized and why they generalize well to data from extensive domains. Especially, he studies the core building blocks of modern deep learning systems and investigates how their training and inference procedures can be made more efficient and reliable. Through this research, Dr. Li has advanced and extended tools from non-convex optimization theory to better explain empirical phenomena observed in practice, and has proposed new theoretical insights that lead to the design of more effective learning algorithms.
Dr. Li was advised by Prof. Meng Wang, Professor of Electrical, Computer, and Systems Engineering, during his time at RPI. “Hongkang is an exceptionally strong and creative researcher who combines deep mathematical rigor with a keen understanding of modern machine learning systems,” says Wang. “His doctoral work addresses fundamental questions about the optimization and generalization of advanced neural networks, including models underlying large language models, and provides principled insights into why these systems perform so well in practice. I believe his research will have a lasting impact on the theory and application of machine learning and signal processing.”
Following his graduation from RPI, Dr. Li is currently a Postdoctoral Researcher in the Department of Electrical and Systems Engineering at the University of Pennsylvania, where he conducts research on the theoretical foundations of the training and generalization mechanisms of deep generative models. As a student at RPI, he was also a recipient of the ML and Systems Rising Star Award from MLCommons in 2023, the Founders Award of Excellence from RPI in 2023, and a Belsky Award for RPI Computational Sciences and Engineering in 2023.