Events for the 5th week of November
-
Center for Systems and Control (CSC@USC) and Ming Hsieh Institute for Electrical Engineering
Mon, Nov 27, 2017 @ 02:00 PM - 03:00 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Maxim Raginsky, University of Illinois at Urbana-Champaign
Talk Title: Decentralized Online Optimization with Global Objectives and Local Communication
Series: Fall 2017 Joint CSC@USC/CommNetS-MHI Seminar Series
Abstract: This talk, based on joint work with Soomin Lee and Angelia Nedich, focuses on a decentralized online convex optimization problem, where each agent controls only one coordinate of the global decision vector. The agents communicate with their neighbors over a static undirected graph or over a time-varying sequence of directed graphs under a uniform connectivity condition. We propose a decentralized variant of Nesterov's primal-dual algorithm with dual averaging. To mitigate the disagreements on the primal-vector updates that arise due to locality of communication, the agents implement a generalization of the local information-exchange dynamics recently proposed by Li and Marden in the undirected case, and a broadcast-based gradient push-sum dynamics in the directed case. We show that, when the step size is chosen appropriately and the objective functions are Lipschitz with Lipschitz gradients, the resulting regret is sublinear in the time horizon.
Biography: Maxim Raginsky received the B.S. and M.S. degrees in 2000 and the Ph.D. degree in 2002 from Northwestern University, all in electrical engineering. He has held research positions with Northwestern, University of Illinois at Urbana-Champaign (where he was a Beckman Foundation Fellow from 2004 to 2007), and Duke University. In 2012, he returned to UIUC, where he is currently an Associate Professor and William L. Everitt Fellow with the Department of Electrical and Computer Engineering. Dr. Raginsky received a Faculty Early Career Development (CAREER) Award from the National Science Foundation in 2013. His research interests lie at the intersection of information theory, machine learning, and control. He is a member of the editorial boards of Foundations and Trends in Communications and Information Theory and IEEE Transactions on Network Science and Engineering.
Host: Ashutosh Nayyar, ashutosh.nayyar@usc.edu
Location: Hughes Aircraft Electrical Engineering Center (EEB) - 132
Audiences: Everyone Is Invited
Contact: Gerrielyn Ramos
-
Center for Cyber-Physical Systems and Internet of Things and Ming Hsieh Institute for Electrical Engineering Joint Seminar Series on Cyber-Physical Systems
Wed, Nov 29, 2017 @ 02:00 PM - 03:00 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Soummya Kar, Associate Professor, Carnegie Mellon University
Talk Title: Resilient Distributed Inference in Cyber-Physical Systems
Abstract: In applications such as large-scale cyber-physical systems (CPS) and Internet-of-Things (IoT), as the number of devices or agents continues to grow, the integrity and trustworthiness of data generated by these devices becomes a pressing issue of paramount importance. An adversary may hijack individual devices or the communication channel between devices to maliciously alter data streams. In numerous IoT applications, we deploy physical devices throughout an environment, and we are interested in using the stream of sensor measurements to make inferences about the environmental state. Due to the large-scale and distributed nature of devices and data it might be infeasible to carry out computation and decision-making in a classical centralized fashion as well as to prevent attacks and intrusions on all data sources. As a result, reactive countermeasures, such as intrusion detection schemes and resilient inference algorithms become a vital component of security in distributed IoT-type setups.
As an alternative to traditional fusion-center based cloud setups, in this talk we focus on fog-type architectures in which devices themselves perform the necessary computations using local data and peer-to-peer information exchange with neighboring devices to make inferences about an environment. In the first part of the talk, we review distributed inference approaches and algorithms based on the consensus+innovations paradigm. We discuss performance metrics such as rates of convergence, communication complexity, and optimality. The second part of the talk focuses on recent work on secure and resilient variants of these algorithms in adversarial environments. Specifically, focusing on the case of data integrity attacks on the device network, we characterize fundamental trade-offs between resilience, quantified in terms of achievable inference performance and ability to detect intrusions and threats, and model properties such as observability and connectivity of the inter-device communication network.
Biography: Soummya Kar received a B.Tech. in electronics and electrical communication engineering from the Indian Institute of Technology, Kharagpur, India, in May 2005 and a Ph.D. in electrical and computer engineering from Carnegie Mellon University, Pittsburgh, PA, in 2010. From June 2010 to May 2011, he was with the Electrical Engineering Department, Princeton University, Princeton, NJ, USA, as a Postdoctoral Research Associate. He is currently an Associate Professor of Electrical and Computer Engineering at Carnegie Mellon University, Pittsburgh, PA, USA. His research interests include decision-making in large-scale networked systems, stochastic systems, multi-agent systems and data science, with applications to cyber-physical systems and smart energy systems. Recent recognition of his work includes the 2016 O. Hugo Schuck Best Paper Award from the American Automatic Control Council and a 2016 Dean's Early Career Fellowship from CIT, Carnegie Mellon.
Host: Paul Bogdan
Location: Hughes Aircraft Electrical Engineering Center (EEB) - 132
Audiences: Everyone Is Invited
Contact: Estela Lopez
-
Memristive Accelerators for Data Intensive Computing: From Machine Learning to High- Performance Linear Algebra
Thu, Nov 30, 2017 @ 02:00 PM - 03:15 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Engin Ipek, University of Rochester
Talk Title: Memristive Accelerators for Data Intensive Computing: From Machine Learning to High- Performance Linear Algebra
Abstract: DRAM is facing severe scalability challenges due to precise charge placement and sensing hurdles in deep-submicron geometries. Resistive memories, such as phase-change memory (PCM), resistive RAM (RRAM), and spin-torque transfer magnetoresistive RAM (STT-MRAM), hold the potential to scale well beyond DRAM and are promising DRAM replacements. Although the near term application of these technologies will likely be in main memory and storage, their electrical properties also make it possible to design qualitatively new methods of accelerating important classes of workloads.
In this talk, I will examine high-performance memristive compute engines that combine two powerful capabilities: in-situ data processing and analog computing. Implementations of these engines using PCM, RRAM, and STT-MRAM will be introduced, and their application to machine learning, combinatorial optimization, and scientific computing workloads will be presented. The talk will conclude with a discussion of the novel error correction techniques that are necessary to make the reliability and precision of memristive accelerators competitive with digital systems.
Biography: Engin Ipek is an Associate Professor of Electrical & Computer Engineering and of Computer Science at the University of Rochester. His research interests are in energy-efficient architectures, high-performance memory systems, and the application of emerging technologies to computer systems. Prof. Ipek received his BS (2003), MS (2007), and Ph.D. (2008) degrees from Cornell University, all in Electrical and Computer Engineering. Prior to joining the University of Rochester, he was a researcher in the computer architecture group at Microsoft Research (2007-2009). His work has been recognized by the 2014 IEEE Computer Society TCCA Young Computer Architect Award, an HPCA 2016 distinguished paper award, three IEEE Micro Top Picks awards, an ASPLOS 2010 best paper award, an NSF CAREER award, and an invited Communications of the ACM research highlights article.
Host: Xuehai Qian, x04459, xuehai.qian@usc.edu
Location: Hughes Aircraft Electrical Engineering Center (EEB) - 132
Audiences: Everyone Is Invited
Contact: Gerrielyn Ramos