2025

Maryam Fazel’s pioneering work on rank-constrained problems and low-rank matrix recovery began with her PhD thesis, where she introduced the now-standard nuclear norm regularization, its SDP representation, and derived key theoretical properties with applications from control to signal processing. Her influential SIAM Review paper with Recht and Parrilo provided the first guarantees for low-rank matrix recovery from surprisingly few measurements, launching a major research direction encompassing matrix and tensor completion, and low-dimensional modeling. With over 4,380 citations and recognition as a 2011 Fast-breaking paper by ScienceWatch, this work stands as both groundbreaking original research and a foundational reference in optimization.

In recent years, Maryam Fazel has advanced approaches to data science and AI, with contributions to meta-learning, online learning, and reinforcement learning. She developed the first theoretical link between flat minima of loss functions and generalization performance of neural networks. She also proposed smoothing-based methods that improve competitive ratios in online algorithms, extending to matrix-valued problems such as optimal experimental design. In reinforcement learning, her optimization-based analysis provided convergence and optimality guarantees for policy optimization, influencing the control community and leading to widely cited work, invited surveys, and tutorials at top conferences.

An outstanding researcher and expositor, Maryam Fazel has delivered numerous plenary and keynote lectures, including at ISMP, the SIAM Annual Meeting, and the American Control Conference. She has served on editorial boards of leading journals as well as in program committees of key conferences in the field, most recently as a co-chair for the International Conference on Machine Learning.