Redirecting to original paper in 30 seconds...

Click below to go immediately or wait for automatic redirect

arxiv_ml 65% Match Research Paper Machine learning researchers,Optimization experts,Data scientists working with distributed systems 2 weeks ago

CoCoA Is ADMM: Unifying Two Paradigms in Distributed Optimization

reinforcement-learning › rlhf
📄 Abstract

Abstract: We consider primal-dual algorithms for general empirical risk minimization problems in distributed settings, focusing on two prominent classes of algorithms. The first class is the communication-efficient distributed dual coordinate ascent (CoCoA), derived from the coordinate ascent method for solving the dual problem. The second class is the alternating direction method of multipliers (ADMM), including consensus ADMM, proximal ADMM, and linearized ADMM. We demonstrate that both classes of algorithms can be transformed into a unified update form that involves only primal and dual variables. This discovery reveals key connections between the two classes of algorithms: CoCoA can be interpreted as a special case of proximal ADMM for solving the dual problem, while consensus ADMM is equivalent to a proximal ADMM algorithm. This discovery provides insight into how we can easily enable the ADMM variants to outperform the CoCoA variants by adjusting the augmented Lagrangian parameter. We further explore linearized versions of ADMM and analyze the effects of tuning parameters on these ADMM variants in the distributed setting. Extensive simulation studies and real-world data analysis support our theoretical findings.
Authors (4)
Runxiong Wu
Dong Liu
Xueqin Wang
Andi Wang
Submitted
February 1, 2025
arXiv Category
math.OC
arXiv PDF

Key Contributions

This paper unifies two prominent paradigms in distributed optimization: CoCoA and ADMM. It demonstrates that both can be transformed into a unified update form, revealing that CoCoA is a special case of proximal ADMM for the dual problem, and consensus ADMM is equivalent to proximal ADMM, providing insights for algorithm improvement.

Business Value

Offers a deeper theoretical understanding of distributed optimization algorithms, potentially leading to more efficient and robust training of large-scale machine learning models used in various industries.