Authors: Fangda Gu*, Jingqi Li*, Ziye Ma*
This is the code repo for EE227C final project.
In the project we introduced two robust methods on the dual:
- [RA-ALM] Robust Accelerated Augmented Laguangian Method
- [RA-ADMM] Robust Accelerated Alternating Direction Method of Multipliers
These methods are runnable using the following scripts:
- [RA-ALM] :
alms/raalm.py
One can find theplot.py
under the same folder which generates the plots. - [RA-ADMM] :
RA_ADMM.m
In addition, the Lyapunov candidate function plot can be generated fromRA_ADMM_Lyapunov.m
.
There are several paper for reference: A Robust Accelerated Optimization Algorithm for Strongly Convex Functions
https://arxiv.org/pdf/1710.04753.pdf
Fast Alternating Direction Optimization Methods
ftp:https://ftp.math.ucla.edu/pub/camreport/cam12-35.pdf
ADMM and Accelerated ADMM as Continuous Dynamical Systems
https://proceedings.mlr.press/v80/franca18a/franca18a.pdf
Accelerated Alternating Direction Method of Multipliers