Second-Order Optimization for Large Language Models
NYU Center for Data Science · Andrew Gordon Wilson
My work explores scaling and improving the efficiency of Gauss–Newton methods as alternatives to first-order optimizers for training large models.
Transient Sprinkler Fluid Dynamics
Applied Mathematics Lab, NYU · Leif Ristroph
Experimental and computational research on real-world fluid dynamics problems, currently investigating the Feynman sprinkler problem.
Smoothing Newton Method for Phase Retrieval
Brandeis University · Tyler Maunu
Researched implicit regularization properties of Newton's Method, targeting quadratic convergence for nonsmooth phase retrieval problems.