Model identification and local linear convergence of coordinate descent
Machine Learning
2020-10-23 v1 Machine Learning
Optimization and Control
Abstract
For composite nonsmooth optimization problems, Forward-Backward algorithm achieves model identification (e.g. support identification for the Lasso) after a finite number of iterations, provided the objective function is regular enough. Results concerning coordinate descent are scarcer and model identification has only been shown for specific estimators, the support-vector machine for instance. In this work, we show that cyclic coordinate descent achieves model identification in finite time for a wide class of functions. In addition, we prove explicit local linear convergence rates for coordinate descent. Extensive experiments on various estimators and on real datasets demonstrate that these rates match well empirical results.
Cite
@article{arxiv.2010.11825,
title = {Model identification and local linear convergence of coordinate descent},
author = {Quentin Klopfenstein and Quentin Bertrand and Alexandre Gramfort and Joseph Salmon and Samuel Vaiter},
journal= {arXiv preprint arXiv:2010.11825},
year = {2020}
}