English

The Randomized Block Coordinate Descent Method in the H\"older Smooth Setting

Optimization and Control 2024-03-14 v1

Abstract

This work provides the first convergence analysis for the Randomized Block Coordinate Descent method for minimizing a function that is both H\"older smooth and block H\"older smooth. Our analysis applies to objective functions that are non-convex, convex, and strongly convex. For non-convex functions, we show that the expected gradient norm reduces at an O(kγ1+γ)O\left(k^{\frac{\gamma}{1+\gamma}}\right) rate, where kk is the iteration count and γ\gamma is the H\"older exponent. For convex functions, we show that the expected suboptimality gap reduces at the rate O(kγ)O\left(k^{-\gamma}\right). In the strongly convex setting, we show this rate for the expected suboptimality gap improves to O(k2γ1γ)O\left(k^{-\frac{2\gamma}{1-\gamma}}\right) when γ>1\gamma>1 and to a linear rate when γ=1\gamma=1. Notably, these new convergence rates coincide with those furnished in the existing literature for the Lipschitz smooth setting.

Keywords

Cite

@article{arxiv.2403.08080,
  title  = {The Randomized Block Coordinate Descent Method in the H\"older Smooth Setting},
  author = {Leandro Farias Maia and David Huckleberry Gutman},
  journal= {arXiv preprint arXiv:2403.08080},
  year   = {2024}
}
R2 v1 2026-06-28T15:17:58.635Z