Full Professor, Zhejiang University
1 paper at NeurIPS 2025
we established optimal risk bound of $1/(\gamma^2 n)$ for GD with deep ReLU networks