An Empirical Study of ADMM for Nonconvex Problems
Optimization and Control
2016-12-13 v1 Machine Learning
Abstract
The alternating direction method of multipliers (ADMM) is a common optimization tool for solving constrained and non-differentiable problems. We provide an empirical study of the practical performance of ADMM on several nonconvex applications, including l0 regularized linear regression, l0 regularized image denoising, phase retrieval, and eigenvector computation. Our experiments suggest that ADMM performs well on a broad class of non-convex problems. Moreover, recently proposed adaptive ADMM methods, which automatically tune penalty parameters as the method runs, can improve algorithm efficiency and solution quality compared to ADMM with a non-tuned penalty.
Cite
@article{arxiv.1612.03349,
title = {An Empirical Study of ADMM for Nonconvex Problems},
author = {Zheng Xu and Soham De and Mario Figueiredo and Christoph Studer and Tom Goldstein},
journal= {arXiv preprint arXiv:1612.03349},
year = {2016}
}
Comments
NIPS nonconvex workshop