Yifan Sun

Optimization and Algorithms

My research centers on the design and analysis of optimization algorithms, particularly those arising in large-scale machine learning and scientific computing. I study how structure — such as sparsity, curvature, or decomposability — can be exploited to design faster, more stable, or more interpretable methods. This includes both first-order methods and quasi-Newton techniques, with applications ranging from convex programming to nonconvex deep models.

Machine Learning Theory

I’m interested in foundational questions in machine learning: generalization, complexity, and the role of structure in model behavior. My work explores how theoretical insights — particularly from linear algebra and geometry — can lead to better understanding of deep models, regularization, and representation learning. I’m also interested in the interplay between optimization and generalization in overparameterized models.