For nonsmooth convex finite-sum optimization, random reshuffling and single shuffle achieve last-iterate rates up to n^{1/4} and n^{1/2} faster than proximal gradient descent, with random reshuffling suffix average matching a known lower bound.
cc/paper_files/paper/2020/file/ 2e2c4bf7ceaa4712a72dd5ee136dc9a8- Paper.pdf
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.OC 1years
2025 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
Improved Last-Iterate Convergence of Shuffling Gradient Methods for Nonsmooth Convex Optimization
For nonsmooth convex finite-sum optimization, random reshuffling and single shuffle achieve last-iterate rates up to n^{1/4} and n^{1/2} faster than proximal gradient descent, with random reshuffling suffix average matching a known lower bound.