[Ml-stat-talks] Princeton Optimization Seminar: Garud Iyengar, Columbia Univ. - TODAY, 4:30 PM

Amir Ali Ahmadi a_a_a at princeton.edu
Thu Apr 30 01:43:40 EDT 2015

-----   Princeton Optimization Seminar   -----

DATE:  Thursday, April 30  (TODAY!)

TIME:  4:30 pm

LOCATION:  Sherrerd Hall 101

SPEAKER: Garud Iyengar, Columbia University

TITLE: A Distributed Proximal Method for Composite Convex Optimization<https://orfe.princeton.edu/abstracts/optimization-seminar/distributed-proximal-method-composite-convex-optimization>

We propose a distributed first-order augmented Lagrangian (DFAL) algorithm for minimizing the sum of composite convex functions, where each term in the sum is only known at one of the nodes, and only nodes connected by an edge can directly communicate with each other. We also allow for global constraints on the solution. This optimization model abstracts a number of applications in distributed sensing and machine learning.
We show that any limit point of DFAL iterates is optimal; and the iterates are epsilon-optimal in O(log(1/epsilon)) iterations, which
require O(ψ3/2max/dmin⋅1/ϵ) communications steps, where psi_max denotes the largest eigenvalue of the graph Laplacian, and dmin is the degree of smallest degree node.
We also propose an asynchronous version of DFAL by incorporating randomized block coordinate descent methods; and demonstrate the efficiency of DFAL on large scale sparse-group LASSO problems.

Garud Iyengar received a B. Tech. in Electrical Engg. from IIT Kanpur in 1993 and a Ph.D. in Electrical Engineering from Stanford University in 1998. Since then he has been with the Department of Industrial Engineering and Operation Research Department at Columbia University where he is currently a Professor. He is the recipient of the President's Gold Medal from IIT Kanpur and the NSF CAREER award. His research interests span a number of different topics such as robust optimization and its applications, large scale optimization for machine learning, information theoretic analysis of biological systems.

If you would like to subscribe to the mailing list of the optimization seminar series, please visit the link below or send an email with
SUBSCRIBE opt-seminar in the body to listserv at lists.princeton.edu<mailto:listserv at lists.princeton.edu>.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/ml-stat-talks/attachments/20150430/ab7ed36b/attachment.html>

More information about the Ml-stat-talks mailing list