SPAdeS-Nets 2013
1st IEEE/ACM Workshop on Signal Processing Advances in Sensor Networks
April 8, 2013, Philadelphia, Pennsylvania, USA (Part of the CPSWeek 2013)
(back to workshop home) PDF program updated Apr. 03, 2013
Organizers
Usman A. Khan, Tufts University
Waheed U. Bajwa, Rutgers, The State University of New Jersey
Workshop Email: spadesnets2013@gmail.com
Workshop Program
08:30am -- 09:15am Ali H. Sayed (UCLA) Abstract, Bio
09:15am -- 09:40am Alejandro Ribeiro (UPenn) Abstract, Bio, Slides
09:40am -- 10:05am Michael Rabbat (McGill) Abstract, Bio, Slides
10:05am -- 10:30am Coffee break
10:30am -- 11:15am José M. F. Moura (CMU) Abstract, Bio
11:15am -- 11:40am Yasamin Mostofi (UCSB) Abstract, Bio, Slides
11:40am -- 12:05pm Ali Tajer (WSU) Abstract, Bio, Slides
12:05pm -- 1:00pm Lunch
01:00pm -- 01:45pm Panel Discussion
02:00pm -- 03:00pm Poster Session (Hotel Foyer) Participants
3:00pm Adjourn
List of abstracts and bios
Inference and Optimization over Networks: It Matters How Information Flows, Ali H. Sayed (UCLA)
Abstract
Adaptive networks consist of a collection of agents with adaptation and learning abilities. The agents interact with each other on a local level and diffuse information across the network to solve estimation, inference, and optimization problems in a distributed manner. Some surprising phenomena arise when information is processed in a decentralized fashion over networks. For example, the collection of more information by the agents is not always beneficial to the inference task and even minor variations in how information is processed by the agents can lead to catastrophic error propagation across the graph. In this talk, we elaborate on such phenomena. In particular, we examine the performance of stochastic-gradient learners for global optimization problems. We consider two well-studied classes of distributed schemes including consensus strategies and diffusion strategies. We quantify how the mean-square-error and the convergence rate of the network vary with the combination policy and with the fraction of informed agents. It will be seen that the performance of the network does not necessarily improve with a larger proportion of informed agents. A strategy to counter the degradation in performance is presented. We also examine how the order by which information is processed by the agents is critical; minor variations can lead to catastrophic failure even when the agents are able to solve the inference task individually on their own. To illustrate this effect, we will establish that diffusion protocols are mean-square stable regardless of the network topology. In contrast, consensus networks can become unstable even if all individual nodes are agents. These results indicate that information processing over networks leads to richer dynamics than originally thought with some revealing learning phenomena.
Bio
Ali H. Sayed (sayed@ee.ucla.edu) is professor and former chairman of electrical engineering at the University of California, Los Angeles, where he directs the UCLA Adaptive Systems Laboratory (http://www.ee.ucla.edu/asl). An author of over 400 scholarly publications and five books, his research involves several areas including adaptation and learning, network science, information processing theories, and biologically-inspired designs. His work received several recognitions including the 2012 Technical Achievement Award from the IEEE Signal Processing Society, the 2005 Terman Award from the American Society for Engineering Education, a 2005 Distinguished Lecturer from the IEEE Signal Processing Society, the 2003 Kuwait Prize, and the 1996 IEEE Fink Prize. He has also been awarded several Best Paper Awards from the IEEE and is a Fellow of both the IEEE and the American Association for the Advancement of Science.
Bayesian Network Games, Alejandro Ribeiro (UPenn)
Abstract
In many situations agents in a network want to take actions corresponding to a joint global operating point while only being able to coordinate and access information in their local neighborhoods. A natural alternative is to let agents select an action that maximizes their expected utility with respect to the information available and a model on the actions they expect will be taken by other agents. We call this action model a Bayesian network game in which agents cooperate with each other while hedging their actions with respect to uncertainty that is inherent in a networked setting. We consider repeated games in which beliefs on the actions of other agents are refined through subsequent observations of neighboring actions. We discuss asymptotic properties of these games and for the particular case of quadratic payoffs we introduce the Quadratic Network Game filter that agents can run locally to update their beliefs, select corresponding optimal actions, and eventually learn a sufficient statistic of the network's state. Open research directions are discussed to close the talk.
Bio
Alejandro Ribeiro received the B.Sc. degree in electrical engineering from the Universidad de la Republica Oriental del Uruguay, Montevideo, in 1998 and the M.Sc. and Ph.D. degree in electrical engineering from the Department of Electrical and Computer Engineering, the University of Minnesota, Minneapolis in 2005 and 2007. From 1998 to 2003, he was a member of the tech- nical staff at Bellsouth Montevideo. After his M.Sc. and Ph.D studies, in 2008 he joined the University of Pennsylvania (Penn), Philadelphia, where he is currently an Assistant Professor at the Department of Electrical and Systems Engineering. His research interests are in the applications of statistical signal processing to the study of networks and networked phenomena. His current research focuses on wireless networks, network optimization, learning in networks, networked control, robot teams, and structured representations of networked data structures. Dr. Ribeiro received the 2012 S. Reid Warren, Jr. Award presented by Penn's undergraduate student body for outstanding teaching and the NSF CAREER Award in 2010. He is also a Fulbright scholar and the recipient of student paper awards at ICASSP 2005 and ICASSP 2006.
Self-Silencing Rules for Randomized Gossip Algorithms, Michael Rabbat (McGill)
Abstract
Randomized gossip algorithms are attractive for collaborative in-network processing and aggregation because they are fully asynchronous, they require no overhead to establish and form routes, and they do not create any bottleneck or single point of failure. Previous studies have focused on analyzing the worst-case number of transmissions required to reach a specified level of accuracy. In a practical implementation, rather than always running for the worst-case number of transmissions, one would like to fix a final level of accuracy and have the algorithm run only until this level of accuracy is achieved, adapting to the initial condition and network topology. I will describe and analyze a local silencing rule for randomized gossip algorithms, whereby nodes locally decide when to stop transmitting in order to conserve transmissions. Theoretical guarantees are provided on the final accuracy of the estimates.
Bio
Michael G. Rabbat earned the B.Sc. degree from the University of Illinois at Urbana-Champaign, in 2001, the M.Sc. degree from Rice University, in 2003, and the Ph.D. from the University of Wisconsin-Madison, in 2006, all in electrical engineering. Currently, he is an Assistant Professor at McGill University, Montreal, Canada. He was a Visiting Researcher at Applied Signal Technology, Inc., during the summer of 2003. Dr. Rabbat's research interests lie in the intersection of statistical signal processing, networking, and machine learning. His current research focuses on distributed signal processing, network inference, and signal processing of data supported on graphs. He received the Best Paper Award (Signal Processing and Information Theory Track) at the 2010 International Conference on Distributed Computing in Sensor Systems (DCOSS 2010), Honourable Mention for Outstanding Student Paper Award at the 2006 Conference on Neural Information Processing Systems (NIPS 2006), and Best Student Paper Award at the 2004 ACM/IEEE International Symposium on Information Processing in Sensor Networks (IPSN 2004). He is currently an Associate Editor for IEEE Signal Processing Letters.
Signal Processing for Graphs , José M. F. Moura (CMU)
Abstract
Data, big or small, in social networks, evolutionary dynamics, the world-wide-web, or citation networks are indexed by social agents, individuals of a population, web sites, or authors all very different from time marks or image pixels. The relations among these data are captured by a graph and not as simple as with data samples in traditional time series, nor pixels in images. We extend traditional discrete signal processing (DSP) tools and concepts to signals defined in graphs.
Bio
Prof. José M. F. Moura was elected University Professor at Carnegie Mellon University to recognize his professional achievement as well as his breadth of interests and competence. This title is conferred on faculty members with exceptional national or international distinction. He was inducted in the National Academy of Engineering in 2013. Prior to joining CMU in 1986, he was on the faculty at Instituto Superior Técnico (IST), the Engineering School of the Technical University of Lisbon (Portugal). He has had visiting faculty appointments at MIT: in 1984-86 as Genrad Associate Professor of Electrical and Computer Engineering (visiting) and in 1999-2000 and 2006-2007 as visiting Professor of Electrical Engineering. He was also a visiting Research Scholar at the University of Southern California in the Summers of 1979-1981. He received his D.Sc. in Electrical Engineering and Computer Science from MIT where he also received his MSc. in Electrical Engineering and the Electrical Engineering degree. He holds a Licenciatura em Engenharia Electrotécnica from IST.
José Moura has been involved as a volunteer in a number of positions with professional societies like IEEE. He is the IEEE Division IX Director Elect (2011) that groups seven IEEE Societies, to serve as IEEE Division IX Director in 2012-13. He currently serves on the IEEE Publications Services and Products Board (PSPB). He was the President of the IEEE Signal Processing Society (SPS) (2008-09) and is currently Past President and Chair of the Nominations and Appointments Committee of the same Society (2010-11). He was a member of the IEEE Educations Activities Board (EAB) (2010) where he chaired the Society Education Outreach Committeee. He was Vice-Chair of the IEEE Publications Services and Products Board (PSPB) (2008). Read his editorials in the IEEE Signal Processsing Magazine (Jan 2008 through Jan 2010) he wrote as President of the Society. He was editor in chief (EIC) for the IEEE Transactions on Signal Processing and acting EIC for the IEEE Signal Processing Letters. He was Vice-President Publications for the IEEE Signal Processing Society (SPS) and Vice-President Publications for the IEEE Sensors Council. He was on the Board of Governors of the IEEE SPS.
José M. F. Moura is a Fellow of the IEEE, a Fellow of AAAS, and a corresponding member of the Academia das Ciéncias de Lisboa (Portugal). He received the 2010 IEEE Signal Processing Society Technical Achievement Award for fundamental contributions to statistical signal processing. He was awarded by CMU the 2008 Philip L. Dowd Fellowship Award for Contributions to Engineering Education and with Prof. Pueschel the 2007 CIT Outstanding Research Award. He received in 2006 an IBM Faculty Award. He was awarded the IEEE Third Millennium Medal and the 2003 IEEE Signal Processing Society Meritorious Service Award.
Co-Optimization of Sensing, Communication, and Navigation in Mobile Networks , Yasamin Mostofi (USCB)
Abstract
The unprecedented growth of sensing, communications, and computation in the past few years has fundamentally changed the way we understand and process information. The vision of a multi-agent robotic network cooperatively learning and adapting in harsh unknown environments to achieve a common goal is closer than ever. In order to realize this vision, however, we need a foundational understanding of the interplay between sensing, communications and control in these systems. On the sensing side, a mobile network tasked with a certain exploratory mission faces an abundance of information. In such an information-rich world, there is simply not enough time to sample the whole environment. On the communication side, the communication between mobile agents can be severely degraded due to several propagation phenomena, making connectivity maintenance challenging. We then have the following important open question: what are the fundamentals of group decision making in these systems, so that the nodes can accurately build an understanding of the environment and accomplish the given task, despite limited sensing and communication? In this talk, I discuss some of our recent work along this line on the co-optimization of sensing, communication and navigation in mobile networks. Inspired by the recent results in non-uniform sampling theory, I show how a robotic network can exploit the sparse transformation of the parameter of interest for cooperative mapping based on only a small number of measurements. More specifically, I show that through proper motion design and by exploiting the sparsity of the map in another domain, it is indeed possible for the network to see through the walls and build a spatial map of occluded obstacles, using only a small number of wireless measurements. To ensure uninterrupted cooperation, I furthermore propose a new approach for communication-aware navigation and control in robotic networks. Along this line, I first show how each robot can assess the channel at unvisited locations through a probabilistic model-discovery and channel-prediction framework. I then show how this stochastic channel learning can be incorporated in robotic path planning, in order to ensure task accomplishment under resource constraints. Overall, our results indicate significant performance improvement over existing approaches.
Bio
Yasamin Mostofi received the BS degree in electrical engineering from Sharif University of Technology, Tehran, Iran, in 1997, and the MS and PhD degrees in the area of wireless communication systems from Stanford University, California, in 1999 and 2004, respectively. She is currently an associate professor in the Department of Electrical and Computer Engineering at the University of California, Santa Barbara. Dr. Mostofi is the recipient of the Presidential Early Career Award for Scientists and Engineers (PECASE), the National Science Foundation (NSF) CAREER award, the IEEE 2012 Outstanding Engineer Award of Region 6, and the Bellcore fellow-advisor award from Stanford Center for Telecommunications among other awards. Her research is on mobile sensor networks. Current research thrusts include communication-aware navigation and decision making in robotic networks, compressive sensing and control, obstacle mapping, robotic routers, and cooperative information processing. She has served on the Control Systems Society conference editorial board since 2008.
Hypothesis Testing Revisited: Theory and Applications, Ali Tajer (WSU)
Abstract
Driven by growing needs for efficient and low-cost data acquisition and inferential processes in large-scale and complex networks (e.g., electricity grids and distributed radars), the design of such networks involves the imposition of advanced cyber layers atop their physical layers. Due to the strong coupling between such physical and cyber layers, effective designs of sensing and inferential algorithms rely fundamentally on exploiting such inherent couplings. This talk describes two aspects of recent advances on hypothesis testing and their roles in sensing and statistical inference in complex networks.
Bio
Ali Tajer received the B.Sc. and M.Sc. degrees in Electrical Engineering from Sharif University of Technology, and the M.A. degree in Statistics and the Ph.D. degree in Electrical Engineering from Columbia University. During 2010-2012 he was a Postdoctoral Research Associate at Princeton University and an Adjunct Assistant Professor at Columbia University. Since August 2012 he has been an Assistant Professor of Electrical and Computer Engineering at Wayne State University. His research interests lie in the general areas of applied statistics, network information theory, and energy systems.
Posters: List of participants
1. Sensor Node Compressive Sampling in Wireless Seismic Sensor Networks
Marc J. Rubin, Michael Wakin, and Tracy Camp
Colorado School of Mines
2. Comparing the Performance of Gossip and Tree-based Protocols for Distributed Averaging
Jun Y. Yu and Michael Rabbat
McGill University
3. Robust Sensor Placement via Graph Sampling under Targeted Neighborhood Attacks
Joya A. Deri and José M. F. Moura
Carnegie Mellon University
4. Multipath Data Aggregation on WSN
Milton A. Cunguara, Tomas A. M. Oliveira e Silva, and Paulo B. R. Pedreiras
University of Aveiro, Portugal
5. Distributed Sequential Estimation of a Linear Model
Yunlong Wang and Petar M.Djuric
Stony Brook University
6. Distributed Topology-Aware Tracking of Multiple Targets
Guohua Ren and Ioannis Schizas
University of Texas at Arlington
7. Compressive Sampling Based High Rate Real-Time Target Monitoring With Low-Power Wireless Sensor Network
Kai Yu , Ming Yin, and Zhi Wang, Zhejiang University
Yu-hen Hu, University of Wisconsin-Madison
8. Toward Resource-Optimal Consensus Over the Wireless Medium
Matthew Nokleby and Robert Calderbank, Duke University
Waheed U. Bajwa, Rutgers University
Behnaam Aazhang, Rice University
9. On the Coherence Properties of Random Euclidean Distance Matrices
Dionysios S. Kalogerias and Athina P. Petropulu
Rutgers University
10. Signal Strength for Sensor Nodes in Two Tier Sensor Networks
Kanwalinderjit K. Gagneja, Kendall E. Nygard
North Dakota State University
11. Measurement-induced topology for distributed estimation
Mohammadreza Doostmohammadian and Usman A. Khan
Tufts University
12. A Study on Strategic Communications in Social Networks
Lin Li, Anna Acaglione, Qing Zhao, UC Davis
Ananthram Swami, Army Research Laboratory
13. Efficient Distributed Algorithms For Structured Optimization Problems
Joao F. C. Mota, CMU
Joao M. F. Xavier and Pedro M. Q. Aguiar, Technical University of Lisbon, Lisbon, Portugal
Markus Pueschel, ETH Zurich
14. Bio-inspired Distributed Time Scheduling in Clustered Networks
Saman Ashkiani and Anna Scaglione
University of California, Davis
15. Coordinated Neighborhood Demand Management for Generation Following
Mahnoosh Alizadeh, Tsung-hui Chang, and Anna Scaglione
University of California Davis
16. Epidemic Process on Fixed Topology Networks
June Zhang and José M. F. Moura
Carnegie Mellon University
17. Robust Decentralized State Estimation and Tracking for Power Systems via Network Gossiping
Xiao Liy and Anna Scaglione
University of California, Davis
18. Sensor Network Field Reconstruction with Time-Varying Target Support
Aurora Schmidt and José M. F Moura
Carnegie Mellon University
School of Engineering
Tufts University