PEOPLE. LUCA SCHENATO. ECC13.

Luca Schenato

Full Professor

Department of Information Engineering

University of Padova

Via Gradenigo 6/B – 35131 Padova Italy

Tel.: +39 049.827.7925

HYCON2 Workshop on Distributed Optimization in Large Networks and its Applications

July 16 July 2013, ETH, Zurich, Switzerland joint with ECC’13 conference

Context and motivations

The International Workshop on Distributed Optimization in Large Networks and its Application is partially supported by the FP7 Framework project HYCON2.

The objective of this workshop is to provide a self-contained overview of this growing body of literature in distributed optimization from a control perspective.

Although several invited sessions and special events have appeared in international conferences, this would be one of the ﬁrst workshop to address the problem of distributed optimization from a control perspective.

More specifically, the proliferation of relatively inexpensive devices capable of communicating, computing, sensing, interacting with the environment and storing information is promising an unprecedented number of novel applications throughout the cooperation of these devices toward a common goal.

These applications include swarm robotics, wireless sensor networks, smart energy grids, smart trafﬁc networks, smart camera networks.

These applications also pose new challenges, of which distributed and asynchronous optimization is one of the major ones. In fact, distributed optimization has being attracting ever growing attention in the past years since many problems in large scale network can be cast as convex optimization problems.

Intended Audience

Registration

Organizers

Organization and program

Schedule

Speakers

First-order methods for distributed in-network optimization

Angelia Nedic

University of Illinois at Urbana-Champaign, USA Abstract: The advances in wired and wireless technology necessitated the development of theory, models and tools to cope with new challenges posed by large-scale optimization problems over networks. We will discuss the current state of these algorithms, as developed recently for distributed multi-agent network systems and their performance properties. The emphasis will be on the interplay between the optimization methods and the network capability to propagate the information. The network effects will be discussed for both synchronous and asynchronous methods including the network connectivity structure, noisy communication links, and other network properties. Some challenging open problems will also be presented. Biography: Angelia Nedich received her B.S. degree from the University of Montenegro (1987) and M.S. degree from the University of Belgrade (1990), both in Mathematics. She received her Ph.D. degrees from Moscow State University (1994) in Mathematics and Mathematical Physics, and from Massachusetts Institute of Technology in Electrical Engineering and Computer Science (2002). She has been at the BAE Systems Advanced Information Technology from 2002-2006. She is currently an Associate professor at the Department of Industrial and Enterprise Systems Engineering at the University of Illinois at Urbana- Champaign, USA, which she joined in fall 2006. She is the recipient of the NSF CAREER Award 2008 in Operations Research for her work in distributed multi-agent optimization. Her general interest is in optimization including fundamental theory, models, algorithms, and applications. Her current research interest is focused on large-scale convex optimization, distributed multi-agent optimization, stochastic approximations in optimization and variational inequalities with applications in signal processing, machine learning, and control.

Decomposition techniques for distributed optimization: a tutorial overview

ADMM and fast gradient methods for distributed optimization

Newton-Raphson consensus for distributed convex optimization

Luca Schenato

University of Padova, Italy Abstract: We address the problem of distributed unconstrained minimization of separable smooth convex cost functions, where the global cost is given by the sum of several local and private costs, each associated to a speciﬁc agent of a given communication network. We speciﬁcally propose a new technique called Newton-Raphson Consensus which is based on linear consensus algorithms. Beside having low computational, communication and memory requirements the technique has also the beneﬁcial properties of requiring very little coordination and naturally supporting time-varying topologies. We show that under suitable assumptions the proposed algorithm guarantees exponential convergence to the global minimum. Finally, we provide some numerical applications and comparisons with alternative strategies available in the literature. Biography: Luca Schenato is currently Associate Professor at the Department of Information Engineering at the University of Padova. He received the Laurea Degree in Electrical Engineering from the University of Padova in 1999, the Management of Technology Certiﬁcate from the Haas Business School at UC Berkeley, USA, in 2003, and the Ph.D. from the Electrical Engineering and Computer Science Department of UC Berkeley in 2003. In 1997 he was a visiting student at the Department of Computing Science at the University of Aberdeen, UK, and in 2004 he help a post-doctoral fellowship at the EECS Department of UC Berkeley. From 2004 to 2007 he was the recipient of the Professorship ”Returning Brains (Rientro dei Cervelli)” sponsored by the Italian Ministry of Education and he served as Adjunct Professor at the Department of Information Engineering at the University of Padova. He won the Eli Jury Award from the EECS Department of UC Berkeley in 2006 for his ”for outstanding achievement in the area of Systems, Communications, Control, or Signal Processing”. His expertise includes distributed control, estiamation and optimization; networked control systems; wireless sensor networks; consensus algorithms; and biomimetic locomotion. He is also serving as Associate Editor for the IEEE Transactions of Automatic Control.

Coordinate descent methods for huge scale problems

Ion Necoara

Politehnica University of Bucharest, Romania Abstract: In this talk we present randomized block coordinate descent methods for minimizing convex optimization problems with linearly coupled constraints over networks and show that they obtain in expectation an ϵ-accurate solution in at most O( N/ϵ ) iterations, where N is the number of nodes in the network. However, the computational complexity per iteration of these methods is much simpler than the one of a method based on full gradient information and each iteration can be computed in a distributed way. We focus on how to choose the probabilities to make these randomized algorithms to converge as fast as possible and we arrive at solving sparse SDPs. Analysis for rate of convergence in probability is also provided. For strongly convex functions we show that these distributed algorithms converge linearly. We also discuss the extension of these algorithms to composite convex optimization and show that similar results hold. Finally, we provide some numerical applications and comparisons with alternative strategies from the literature. Biography: Ion Necoara received the B.Sc. degree in mathematics and the M.Sc. degree in optimization and statistics from the University of Bucharest, in 2000 and 2002. After graduating he worked as a Ph.D. student at the Delft Center for Systems and Control, Delft University of Technology, The Netherlands. From August to November 2005 he was an invited researcher at University of Cambridge, UK. For the period 2006–2009, he completed a Postdoctoral Fellowship at the Katholieke Universiteit Leuven, Belgium. Since 2009 he is a staff member of the Faculty of Automatic Control and Computers, University Politehnica Bucharest, where he is now an Associate Professor of Numerical Methods and Optimization. His main research interests cover various topics in developing new optimization algorithms with a focus on structure exploiting and applying optimization techniques for developing new advanced controller design algorithms for complex systems.

PEOPLE. LUCA SCHENATO. ECC13.

Luca Schenato

Full Professor

Menu

> ECC13

HYCON2 Workshop on Distributed Optimization in Large Networks and its Applications

July 16 July 2013, ETH, Zurich, Switzerland joint with ECC’13 conference

Context and motivations

Intended Audience

Registration

Organizers

Organization and program

Schedule

Speakers

First-order methods for distributed in-network optimization

Angelia Nedic

Decomposition techniques for distributed optimization: a tutorial overview

ADMM and fast gradient methods for distributed optimization

Newton-Raphson consensus for distributed convex optimization

Coordinate descent methods for huge scale problems

Industrial applications of large-scale and distributed optimization

Menu

> HYCON2

Test