Reproduce “Optimal Stochastic Non-smooth Non-convex Optimization through Online-to-Non-convex Conversion” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ashok Cutkosky, Harsh Mehta, Francesco Orabona
The authors reduce stochastic optimization of non-smooth, non-convex objectives to online learning, targeting a finite, balanced neighborhood whose average gradient vector is small. Random gradient queries along update segments connect objective change to online regret. The method improves the prior query complexity from O(ε⁻⁴δ⁻¹) to O(ε⁻³δ⁻¹); a matching dependence is proved only in a stated parameter regime.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.