Notice: This lecture is a research seminar for credit.
Time
————————————
Mar 28, 2019 9:00-10:00 a.m.
Abstract
————————————
We focus on solving a distributedconvex optimization problem in a network, where each agent has its own convexcost function and the goal is to minimize the sum of the agents‘ cost functionswhile obeying the network connectivity structure. In order to minimize the sumof the cost functions, we consider a new distributed gradient-based methodwhere each node maintains two estimates, namely, an estimate of the optimaldecision variable and an estimate of the gradient for the average of theagents’ objective functions. From the viewpoint of an agent, the informationabout the gradients is pushed to the neighbors, while the information about thedecision variable is pulled from the neighbors hence giving the name “push-pullgradient methods”. The method unifies the algorithms with different types ofdistributed architecture, including decentralized (peer-to-peer), centralized(master-slave), and semi-centralized (leader-follower) architecture. We showthat the algorithm converges linearly for strongly convex and smooth objectivefunctions over a directed static network. In our numerical test, the algorithmperforms well even for time-varying directed networks.
Speaker's Bio
————————————
Shi Pu is a postdoctoralassociate in the Division of Systems Engineering at Boston University. Hereceived a B.S. Degree in Engineering Mechanics from Peking University, in2012, and a Ph.D. Degree in Systems Engineering from the University ofVirginia, in 2016. He was a postdoctoral associate at the University of Floridafrom 2016 to 2017, and a postdoctoral scholar at Arizona State University from2017 to 2018. His research interests include distributed optimization,large-scale data analytics, network science and machine learning.
Registration
————————————
Professors and students of TBSI are welcometo attend. The lecture is also open to the public. For off-campuspersonnel, please scan the QR code and and fill in your information (name,company, contact number, ID number). The language of the lecture isEnglish.
