generalized cross-validation for right-rotationally invariant designs (arxiv, thesis)
- my senior thesis under Pragya Sur and Yufan Li; AISTATS 2025
- we generalize results from [Hastie et al. '20] to right
rotationally invariant design. in particular, we find that generalized cross-validation is no longer
consistent for the out-of-sample loss, and we develop a consistent alternative.
pruning neural networks at initialization (arxiv)
- with Tanishq Kumar, Mark Sellke; ICML 2024
- We prove a modification of the Law of Robustness stated in [Bubeck-Sellke '22] with implications for pruning neural
networks at initialization.
witness encryption demo (github, arxiv)
- with Gavin Uberti, Wittmann Goh, Oliver Cheng
- group final project for cs 227