Rosie Zhao

Hi there! I’m a second year PhD student in Computer Science at Harvard University advised by Boaz Barak and Sham Kakade. I’m interested in reinforcement learning, language models, and broadly establishing machine learning foundations. I previously completed my undergrad and Master’s at McGill University under the supervision of Prakash Panangaden; there, my research was geared towards representation learning and state abstractions in RL, as well as various topics in theoretical computer science.

My socials can be found below.

news

Nov 13, 2023 New preprint on explaining why “Fourier features” emerge in neural networks when learning modular addition (and composition for finite groups), using max-margin analysis! In collaboration with Depen Morwani, Ben Edelman, Costin-Andrei Oncescu, and Sham Kakade. :zap:
Oct 10, 2023 Started as a visiting researcher at ServiceNow Research in Montreal working with Alexandre Piché at the intersection of RL and LLMs!
Aug 25, 2023 Work on Loss of Plasticity in Deep Continual RL was accepted as an oral to CoLLAs 2023! In collaboration with some wonderful researchers at Deepmind Alberta: Zaheer Abbas, Joseph Modayil, Adam White, and Marlos Machado.
Jun 14, 2023 New preprint on the insignificance of SGD noise in online (single-epoch) learning! Done in collaboration with Nikhil Vyas, Depen Morwani, Gal Kaplun, Boaz Barak, and Sham Kakade. :zap: