Associate Professor, Shanghai Jiao Tong University
1 paper at NeurIPS 2025
We propose an adaptive, parameter-free accelerated method for local Hölder smooth convex optimization with optimal rates and no need for Lipschitz or target accuracy tuning.