Projects

Selected coursework projects from Berkeley.

A Mirror-Descent View of Saddle-to-Saddle Dynamics in Deep Linear Networks

Divit Rawal

CS 294 (Learning for Dynamics and Control), Spring 2026

[pdf]

Predicting Benign Overfitting via Spectral Geometry

Divit Rawal

CS 281A (Statistical Learning Theory), Fall 2025

[pdf] / [poster]

Spectrum-Adaptive Randomized SVD for Compressed Retrieval

Divit Rawal, Aneesh Durai

STAT 260 (Randomized Linear Algebra, Optimization, and Large-Scale Learning), Spring 2025

[pdf]