Spring 2015 Seminar Series

DATE: 5/14/15
LOCATION: E51-395
TIME: 4:15pm
Reception immediately following

SPEAKER:
Asu Ozdaglar

TITLE
Distributed Alternating Direction Method of Multipliers for Multi-agent Optimization

ABSTRACT
We consider a network of agents solving a global optimization problem, where the objective function is the sum of privately known local convex objective functions. Recent literature presented subgradient based methods for distributed solution of this problem with O(1/ \sqrt{k}) rate of convergence (where k is the iteration number). In this talk, we present distributed Alternating Direction Method of Multipliers (ADMM) based methods for solving this problem over undirected and directed networks. We present convergence rate estimates that show that these methods converge at rate $O(1/k)$ and highlight the dependence of performance on network structure. We further assume that the local objective functions are strongly convex and have Lipschitz continuous gradients and show that the sequence generated by the method converges linearly to the optimal solution.