News

At the 2026 International Conference on Learning Representations (ICLR), NYU Courant CS student Noah Amsel, Courant Instructor David Persson, and Assistant Professor Christopher Musco, along with their collaborator Robert M. Gower, received an Outstanding Paper Honorable Mention. Their paper, "The Polar Express: Optimal Matrix Sign Methods and Their Application to the Muon Algorithm," addresses fundamental efficiency challenges in deep learning optimization. The research develops a minimax-optimal approach to computing matrix sign functions, which are critical to the Muon optimizer. By reducing the number of matrix multiplications required for convergence and improving stability in low-precision numerical formats like bfloat16, their work provides a more computationally efficient framework for training large-scale neural networks.

Read here