University of South Carolina

Short Course Introduction

ADMM is a very popular and powerful algorithm in big data research area. This is the course webpage of a two-lecture short course designed to make "everybody" understand the ADMM in both theoretical and practical manner.

In lecture one, I plan to cover some basic concepts in convex optimization problems so that students have sufficient building blocks to use when dealing with optimization algorithms. Pictures and examples are offered in the explanation.

In lecture two, I plan to introduce some optimization algorithms including ADMM, and finish the course with concrete examples, i.e., solving LASSO problem with ADMM.

All teaching materials are coming from the listed references. Concepts in lecture one is extracted from the first five chapters of the book "Convex Optimization" by S. Boyd and L. Vandenberghe. Algorithms in lecture two are mainly from the 2010 paper by S. Boyd et al. Book "Convex Analysis" by T. Rockafellar (1970) is called the "holly bible" in the convex analysis field. It is helpful when there is anything unfamiliar or blurred, i.e. subgradients.

Lecture Notes

3/15/2016: Theoretical Tuesday
3/17/2016: Algorithmic Thursday

References

1. Convex Optimization (S. Boyd and L.Vandenberghe)
2. Convex Analysis (T. Rockafellar)
3. Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers (S. Boyd et al.)