Block I logo
University of Illinois Urbana-Champaign
The Grainger College of Engineering
Coordinated Science Laboratory
Search
  • About

    About

    • Celebrating 70 Years of Science
    • Job Opportunities
    • Traveling to CSL
    • CSL 2021 Strategic Plan and Visioning
  • Research

    Research

    • Impact Areas
      • Cybersecurity
        • Data science and cyber trust
        • Quantifying vulnerabilities
        • Secure systems and networking
      • Health IT
        • Cyber-physical-human systems
        • Data science
        • Machine-to-machine communications
        • Novel computing platforms and information processing approaches
        • Precision Medicine
        • Real-time processing and networking
        • Sensors and signal processing
        • Trustworthiness
      • Internet of Things
      • Robotics
        • Collaborative robotic teams
        • Mobile Robotics
        • Robotic intelligence
    • Groups
      • Circuits
      • Computational and Physical Electronics
      • Computer Systems and Architecture
      • Decision and Control
        • Calendar
        • Past Seminars
      • Health Care Engineering Systems Center
      • Intelligent Sensing, Networking, and Cyberphysical Systems
      • Remote Sensing and Space Science
      • Robotics
      • Security & Privacy
      • Signals, Inference, and Networks
      • Technology and Societal Impacts
    • Centers
      • Advanced Electronics through Machine Learning (CAEML)
      • Center for Autonomy
      • Center for Networked Intelligent Components and Environments (C-NICE)
      • Comp. Biotechnology and Genomic Med (CCBGM)
      • Health Care Engineering Systems Center (HCESC)
      • IBM-Illinois Discovery Accelerator Institute
      • Internet of Battlefield Things (IoBT)
      • National Center for Principled Leadership & Research Ethics (NCPRE)
        • People
    • Institutes
      • Advanced Digital Sciences Center
      • Information Trust Institute
    • Initiatives
  • News and Media

    News and Media

    • News Stories
    • Connect Newsletter
    • Videos
    • Annual Reports
    • Uplink Newsletter
    • Tech Reports
  • Events

    Events

    • Conferences
    • CSL Compute-Energy-Nexus Workshop
    • CSL PhD Thesis Award
    • CSL Social Hour
    • Full Calendar
    • Future of Computing Symposium
    • Instarecon Innovation Scholarship
    • Lectures
      • CSL Distinguished Lecture Series
      • CSL Spotlight Lecture Series
      • Robert T. Chien Distinguished Lecturer Series
    • The Symposium on Artificial Intelligence and Social Responsibility
  • Directory

    Directory

    • Faculty
    • Staff
    • Graduate Students
    • Postdocs
    • Visiting Scholars
    • Business and Technical Contacts
  • Contact Us
  • Home
  • Research
  • Groups
  • Decision and Control
  • Calendar
In This Section
  • Research Overview
  • Impact Areas
    • Cybersecurity
      • Data science and cyber trust
      • Quantifying vulnerabilities
      • Secure systems and networking
    • Health IT
      • Cyber-physical-human systems
      • Data science
      • Machine-to-machine communications
      • Novel computing platforms and information processing approaches
      • Precision Medicine
      • Real-time processing and networking
      • Sensors and signal processing
      • Trustworthiness
    • Internet of Things
    • Robotics
      • Collaborative robotic teams
      • Mobile Robotics
      • Robotic intelligence
  • Groups
    • Circuits
    • Computational and Physical Electronics
    • Computer Systems and Architecture
    • Decision and Control
      • Calendar
      • Past Seminars
    • Health Care Engineering Systems Center
    • Intelligent Sensing, Networking, and Cyberphysical Systems
    • Remote Sensing and Space Science
    • Robotics
    • Security & Privacy
    • Signals, Inference, and Networks
    • Technology and Societal Impacts
  • Centers
    • Advanced Electronics through Machine Learning (CAEML)
    • Center for Autonomy
    • Center for Networked Intelligent Components and Environments (C-NICE)
    • Comp. Biotechnology and Genomic Med (CCBGM)
    • Health Care Engineering Systems Center (HCESC)
    • IBM-Illinois Discovery Accelerator Institute
    • Internet of Battlefield Things (IoBT)
    • National Center for Principled Leadership & Research Ethics (NCPRE)
      • People
  • Institutes
    • Advanced Digital Sciences Center
    • Information Trust Institute
  • Initiatives

Angelia Nedich - Tutorial Session - 57th Allerton Conference on Communication, Control, and Computing

Event Type
Conference/Workshop
Sponsor
Coordinated Science Laboratory
Location
CSL Auditorium, Room B02
Date
September 24, 2019 2:00 PM - 5:30 PM
Speaker
Professor Angelia Nedich, Arizona State University
Cost
$40 advanced registration (before Sept 9), $50 registration fee (Sept 10-27) - Contact Peggy Wells pwells@illinois.edu if you have an UIUC affiliation
Registration
Registration
Contact
Angie Ellis
Email
amellis@illinois.edu
Phone
217-300-1910

SCHEDULE

Registration: 1:00-2:00 pm in the lower level lobby of Coordinated Science Laboratory

Afternoon Tutorial by Angelia Nedich 
Session begins: 2:00pm-3:30pm
Afternoon break: 3:30pm-4:00pm (Snacks and water will be provided)
Session resumes: 4:00pm-5:30pm

Conference Welcome Reception will follow from 6:00-8:00pm in the Electrical and Computer Engineering Building, Room 3002

Angelia Nedich will start her tutorial presentation at 2:00pm-5:30pm (includes a 30 minute break from 3:30-4:00pm)

Title: Distributed Algorithms for Optimization in Networks

Abstract:

We will overview the distributed optimization algorithms starting with the basic underlying idea illustrated on a prototype problem in machine learning. In particular, we will focus on convex minimization problem where the objective function is given as the sum of convex functions, each of which is known by an agent in a network. The agents communicate over the network with a task to jointly determine a minimum of the sum of their objective functions. The communication network can vary over time, which is modeled through a sequence of graphs over a static set of nodes (representing the agents in a system). In this setting, the basic distributed first-order methods will be discussed that make use of a consensus protocol, which is a mechanism virtually replacing the role of a coordinator. In these basic distributed methods, the agents are somewhat selfish as they greedily process only the gradients of their own objective function, while they mix (through a consensus step) their decision variables. As a result, the methods are slow and cannot be used with a constant stepsize value, which is desirable for a faster convergence. Nevertheless, these methods are useful in the implementations with noisy (stochastic) gradients or noisy/unreliable communication links.

We will then discuss some recent developments of fast distributed gradient methods that can match the performance of the centralized gradient method. In particular, these fast methods are constructed by having the agents be aware of the system objective function in the sense that they are more cooperative in the process of diffusing the gradient information, while still directly processing only their own local objective function. The diffusion of the gradient information is based on a gradient-tracking mechanism, which allows for a proper mixing of the gradient-based directions that the agents are using. Through the diffusion of the agent decision variables and directions, the distributed methods with gradient tracking work with a suitably selected constant stepsize and can be as fast as a centralized gradient method. One of these methods (Apush-Bpull) recently developed, surprisingly, uses a weighted-averaging consensus to diffuse the decision variables, and a push-sum consensus to diffuse the directions. This is quite interesting since most of the prior work on the distributed methods with gradient tracking have used either weighted-averaging consensus or push-sum consensus depending on whether the underlying connectivity graph is undirected or directed. Finally, we will discuss the convergence rate results for these fast distributed methods, provide some intuition behind their analysis, and provide some simulation results that demonstrate their performance.

Biography: Angelia Nedich has a Ph.D. from Moscow State University, Russia, in Computational  Mathematics and Mathematical Physics (1994), and a Ph.D. from Massachusetts Institute of Technology, Cambridge, USA in Electrical and Computer Science Engineering (2002). She is a professor at the school of Electrical, Computer and Energy Engineering at Arizona State University (ASU) at Tempe. Prior to joining ASU, she has been a Willard Scholar faculty at the University of Illinois, Urbana-Champaign. She is a recipient (jointly with co-authors) of the Best Paper Awards at the Winter Simulation Conference 2013, and the International Symposium on Modeling and Optimization in Mobile, Ad Hoc and Wireless Networks 2015. Her research interest is in optimization theory and algorithms, stochastic optimization, variational inequality problems, Nash equilibrium problems, and dynamics of large scale complex systems.

Coordinated Science Laboratory

  • Twitter
  • Instagram
  • Facebook
  • LinkedIn
  • YouTube

CSL Building

1308 W Main Street MC 228

Urbana, IL 61801

Phone: 217-333-2511

Fax:

Email:

The Grainger College of Engineering

Contact Us

  • Privacy Policy
  • Copyright © 2025
  • Accessibility
  • Webmaster
  • Login