2011 Poster Sessions : Distributed Optimization

Student Name : Eric Chu, Arezou Keshavarz, Brendan O'Donoghue
Advisor : Stephen Boyd
Research Areas: Information Systems
Abstract:
Recent trends in Big Data pose a challenge for optimization algorithms. We present techniques to perform distributed optimization, such as dual decomposition and the alternating direction method of multipliers (ADMM). In particular, ADMM is able to bootstrap current solvers to solve problems on an immense scale. This algorithm can be run on a cluster or cloud (e.g., Amazon EC2) and can be used to solve various problems such as classification and regression in machine learning.

Bios:
Arezou Keshavarz, Brendan O'Donoghue, and Eric Chu are Ph.D. students working with Prof. Stephen Boyd. They work on convex optimization and it's applications to machine learning, signal processing, control, finance, statistics and many others.