Arizona State University Network Science Seminar Series

Energy-Efficient Node Deployment in Heterogeneous Sensor Networks

Speaker Hamid Jafarkhani (University of California, Irvine)
Date 2:00 pm, on October 4, 2019
Location GWC 487
Short Bio
Hamid Jafarkhani is a Chancellor's Professor at the Department of Electrical Engineering and Computer Science, University of California, Irvine, where he is also the Director of Center for Pervasive Communications and Computing and the Conexant-Broadcom Endowed Chair. He was a Visiting Scholar at Harvard University in 2015 and a Visiting Professor at California Institute of Technology in 2018. Among his awards are the IEEE Marconi Prize Paper Award in Wireless Communications, the IEEE Communications Society Award for Advances in Communication, and the IEEE Eric E. Sumner Award. Dr. Jafarkhani is listed as an ISI highly cited researcher. According to the Thomson Scientific, he is one of the top 10 most-cited researchers in the field of "computer science" during 1997-2007. He is the 2017 Innovation Hall of Fame Inductee at the University of Maryland's School of Engineering. He is a Fellow of AAAS, an IEEE Fellow, and the author of the book "Space-Time Coding: Theory and Practice."
Abstract
Wireless networks of the future are envisioned to be highly heterogeneous. In many applications, one is interested in optimally deploying a network of nonidentical nodes to a certain area of interest. These networks may include a multitude of connected autonomous nodes in one or more tiers. We formulate these deployment problems as quantizer design problems where different distortion measures should be associated with different quantization indices. We discuss fundamental design challenges like the best spatial deployment of nodes to minimize the energy consumption or maximize the sensing accuracy while guaranteeing network connectivity. This is done by developing a quantization theory of heterogeneous reproduction points. We will discuss the characteristics of the heterogeneous networks with optimal deployment.

Distributed Learning with Gossip Algorithms

Speaker Michael Rabbat (Facebook)
Date 3:00 pm, on October 3, 2019
Location GWC 487
Short Bio
Michael Rabbat a Research Scientist in Facebook's Artificial Intelligence Research group (FAIR) in Montreal, Canada. He received the B.Sc. from the University of Illinois, Urbana-Champaign, in 2001, the M.Sc. from Rice University, Houston, TX, in 2003, and the Ph.D. from the University of Wisconsin, Madison, in 2006, all in electrical engineering. From 2007 to 2018 he was a professor in the Department of Electrical and Computer Engineering at McGill University. During the 2013--2014 academic year he held visiting positions at Telecom Bretegne, France, the Inria Bretagne-Atlantique Research Centre, France, and KTH Royal Institute of Technology, Sweden. His research interests include optimization and distributed algorithms for machine learning.
Abstract
Distributed data-parallel algorithms aim to accelerate the training of deep neural networks by parallelizing the computation of large mini-batch gradient updates across multiple nodes. Approaches that synchronize nodes using exact distributed averaging (e.g., via AllReduce) are sensitive to stragglers and communication delays. The PushSum gossip algorithm is robust to these issues, but only performs approximate distributed averaging. In this talk I will discuss our recent work studying Stochastic Gradient Push (SGP) for supervised learning and Gossip-Based Actor Learner Architectures (GALA) for reinforcement learning, both of which build on PushSum. By reducing the amount of synchronization between compute nodes, both methods are more computationally efficient and scalable compared to comparable methods built on AllReduce, and both methods also enjoy theoretical guarantees, e.g., related to convergence (for SGP) and bounded disagreement (for SGP and GALA). The talk is based on joint work with Mido Assran, Nicolas Ballas, Nicolas Loizou, Josh Romoff, and Joelle Pineau.

Power Quality Data Analytics and Applications

Speaker Surya Santoso (The University of Texas at Austin)
Date 10:00 am, on September 5, 2019
Location GWC 487
Short Bio
Surya Santoso (F’15) earned his B.S. degree from Satya Wacana Christian University, Salatiga, Indonesia, in 1992, and M.S.E. and Ph.D. degrees in Electrical and Computer Engineering from The University of Texas at Austin, in 1994 and 1996, respectively. He was a Senior Power Systems and Consulting Engineer with Electrotek Concepts, Knoxville, TN, USA, from 1997 to 2003. He joined the faculty of The University of Texas at Austin in 2003 and is currently Professor of Electrical and Computer Engineering. His research interests include power quality, power systems, and renewable energy integration in transmission and distribution systems. He is co-author of Electrical Power Systems Quality (3rd edition), sole author of Fundamentals of Electric Power Quality, and editor of Handbook of Electric Power Calculations (4th edition) and Standard Handbook for Electrical Engineers (17th edition). He is an IEEE Fellow.
Abstract
Power quality measurement data are voluminous. They are collected by power quality monitors and intelligent electronic devices installed in distribution substations, feeders, and service entrances. Unfortunately, the task of converting data to the knowledge of situational awareness of equipment and system-level conditions has been often performed manually. This presentation describes efforts to automate and mimick manual human thought process in analyzing voluminous power quality data. Several applications will be shown: identification of operation of protective devices and their coordination, detecting incipient cable faults and predicting its locations, and health conditions of capacitor banks and their switching devices.

Controlling Machine Learning For Regularization, Interpretability, and Fairness

Speaker Maya Gupta (Google AI)
Date 11:15 am, on August 28, 2019
Location GWC 487
Short Bio
Maya Gupta is a Principal Scientist at Google AI, and leads the Glassbox Machine Learning R&D team, which focuses on designing and developing controllable and debuggable machine learning algorithms that solve Google product needs. Prior to Google, Gupta was an Associate Professor of Electrical Engineering at the University of Washington from 2003-2013, where she was awarded the PECASE and ONR YIP awards. Her Ph.D. is in Electrical Engineering from Stanford University (2003), where she was a National Science Foundation Graduate Fellow and worked with Bob Gray. She holds a BS in Electrical Engineering and BA in Economics from Rice University. Maya has also worked for research groups at NATO, Ricoh, AT&T, and HP, and she runs Artifact Puzzles, the second largest US maker of wooden jigsaw puzzles, a company she founded in 2009.
Abstract
What tools can we use to control machine learning, beyond how we sample training data? We will discuss two useful tools, shape constraints and rate constraints. We will show how shape constraints can capture domain knowledge to better regularize models even with challenging distribution shifts of the test data, and impose deontological individual fairness policies, even for functions on sets of inputs. We will show how rate constraints can be used to achieve better metrics for precision at recall or fairness, and present some state-of-the-art algorithms for the resulting constrained optimization problems that borrow from game theory. We will discuss some of the problems and future challenges for controlling ML.

Robust Methods for Influencing Strategic Behavior

Speaker Philip Brown (University of Colorado at Colorado Springs)
Date 3:00 pm, on March 28, 2019
Location GWC 487
Short Bio
Philip Brown is an Assistant Professor in the Department of Computer Science at the University of Colorado at Colorado Springs. He received the PhD in Electrical and Computer Engineering from the University of California, Santa Barbara under the supervision of Jason Marden. He received the Master and Bachelor of Science in Electrical Engineering from the University of Colorado at Boulder and Georgia Tech (respectively), between which he developed process control technology for the biofuels industry. Philip is interested in the impact of human social behavior on the performance of large-scale infrastructure and software systems, and studies this by combining concepts from game theory and feedback control of distributed systems. Philip was a finalist for best student paper at IEEE CDC in 2016 and 2017.
Abstract
The web of interconnections between today's technology and society is upending many traditional ways of doing things: the internet of things, bitcoin, the sharing economy, and stories of "fake news"; spreading on social media are increasingly in the public mind. As such, computer scientists and engineers must be increasingly conscious of the interplay between the technical performance of their systems and the personal objectives of users, customers, and adversaries. I will present work on three problem domains: robust incentive design for smart cities and transportation systems, resilient coordination and optimization in distributed multiagent engineered systems, and how malicious entities can spread their influence in society. In each, by rigorously examining the fundamental tradeoffs associated with optimal designs, I seek to develop a deeper theoretical understanding of tools which will help address today's emerging challenges.

Bayesian learning with non-myopic strategic agents

Speaker Achilleas Anastasopoulos (EECS, University of Michigan, Ann Arbor)
Date 10 am, on March 4, 2019
Location GWC 487
Short Bio
Achilleas Anastasopoulos received the Diploma in EE from the National Technical University of Athens, Greece in 1993, and the M.S. and Ph.D. degrees in EE from University of Southern California in 1994 and 1999, respectively. He is currently an Associate Professor at the University of Michigan, Ann Arbor, Department of EECS. His research interests lie in the general area of resource allocation and information elicitation on networked systems with strategic agents, as well as the connections between stochastic control, communications and information theory.
Abstract
We consider problems involving multiple strategic (selfish) agents making decisions dynamically in the presence of asymmetric information. Specifically, we consider an environment where many players need to decide whether to buy a certain product (or adopt a technology) or not. The true value of the product is not known to the players; instead, each player observes previous players' actions and has his own noisy private information on the product quality. It is well known that in such settings, agents start ignoring their private information thus generating trends/fads, also known as informational cascades, where learning stops in the network as a whole. These results, however, depend on a crucial assumption: players only enter the marketplace once and so they can act myopically. What happens when agents are given the option to return to the market? Agents can no-longer act myopically and have to strategize over the entire time horizon. Clearly there are tradeoffs between waiting for more information and buying early (due to discounting).
In this talk, we present a new methodology for characterizing "structured" Perfect Bayesian Equilibria (SPBE) of these dynamic games with asymmetric information akin to the backward sequential decomposition in MDPs and POMDPs. The corresponding "state" is an appropriate belief based on the common information among agents. By applying this methodology to the problem at hand, and by identifying sufficient statistics that summarize the "state", we provide a characterization of SPBE with non-myopic strategies through a fixed-point equation of dimensionality that grows only quadratically with the number of players. Based on this characterization we study informational cascades and show that they occur with high probability. Furthermore, only a small portion of the total information in the system is revealed before a cascade occurs.

Random Access with Energy Harvesting Nodes

Speaker Tolga M. Duman (Bilkent University)
Date 3:00 p.m., Feb 20, 2019
Location GWC 487
Short Bio
Tolga M. Duman is a Professor of Electrical and Electronics Engineering Department at Bilkent University in Turkey. He received the B.S. degree from the same university in 1993, M.S. and Ph.D. degrees from Northeastern University, Boston, MA, in 1995 and 1998, respectively, all in electrical engineering. Prior to joining Bilkent University, he has been with the Electrical Engineering Department of Arizona State University. Dr. Duman's current research interests are in systems, with particular focus on communications and signal processing, including wireless and mobile communications, coding/modulation, coding for wireless communications, data storage systems and underwater acoustic communications. Dr. Duman is a Fellow of IEEE, a recipient of the National Science Foundation CAREER Award and IEEE Third Millennium medal. He has served as an editor for various journals, and he is currently the coding and information theory area editor of IEEE Trans. on Communications and the Editor-in-Chief of Elsevier's Physical Communication.
Abstract
We propose an irregular repetition slotted ALOHA (IRSA) based uncoordinated random access scheme for energy harvesting (EH) nodes. Specifically, we consider the case in which each user has a battery that is recharged with harvested energy from the environment in a probabilistic manner. We analyze this scheme starting with a unit-sized battery at the nodes and extend the analysis to the case of finite-sized battery. For both scenarios, we derive asymptotic throughput expressions, and obtain optimized probability distributions for the number of packet replicas for the users. We demonstrate that for the case of IRSA with EH nodes, these optimized distributions perform considerably better than the alternatives, especially in the limited battery scenarios, including slotted ALOHA (SA), contention resolution diversity slotted ALOHA (CRDSA) and IRSA, which do not take into account the EH process for both asymptotic and finite frame length scenarios. We also present some extensions of the asymptotic analysis to the contention resolution ALOHA without any slot synchronization.

Seminars

Title Speaker Time Location
Energy-Efficient Node Deployment in Heterogeneous Sensor Networks Hamid Jafarkhani (University of California, Irvine) 2:00 p.m., October 4, 2019 GWC 487
Distributed Learning with Gossip Algorithms Michael Rabbat (Facebook) 3:00 p.m., October 3, 2019 GWC 487
Power Quality Data Analytics and Applications Surya Santoso (The University of Texas at Austin) 10:00 a.m., September 5, 2019 GWC 487
Controlling Machine Learning For Regularization, Interpretability, and Fairness Maya Gupta (Google AI) 11:15 a.m., August 28, 2019 GWC 487
Robust Methods for Influencing Strategic Behavior Philip Brown (University of Colorado at Colorado Springs) 3:00 p.m., March 28, 2019 GWC 487
Bayesian learning with non-myopic strategic agents Achilleas Anastasopoulos (EECS, University of Michigan, Ann Arbor) 10:00 a.m., March 4, 2019 GWC 487
Random Access with Energy Harvesting Nodes Tolga M. Duman (Bilkent University) 3:00 p.m., Feb 20, 2019 GWC 487
 
��ҳģ����ҳģ��