Seminars

2015

Date
Time
Lecturer
Affiliation
Topic
Resources
Top of page ↑
January
1/29
4:30PM
Dr. Somayeh Sojoudi
New York University
Computational Methods for the Analysis and Modeling of Brain Networks
February
2/05
4:30PM
David Z. Pan
University of Texas at Austin
2/11
3:0PM
William Stanchina
University of Pittsburgh
2/12
4:30PM
Dr. Eylem Ekici
Ohio State University
2/19
4:30PM
Dr. Morgado Dias
University of Madeira
Multi-Camera Synchronisation for the NanEye CMOS Image Sensor
2/26
4:30PM
Namrata Vaswani
Iowa State
ABSTRACT: This work studies the problem of sequentially recovering a time sequence of sparse vectors x_t and vectors from a low-dimensional subspace l_t from knowledge of their sum m_t:=x_t+l_t at each time t. The subspace of l_t's changes slowly enough so that the matrix L_t:= [l_1, l_2, ... l_t] is low-rank at each time t. Clearly the matrix X_t:=[x_1, x_2, ... x_t] is sparse. Thus this is a problem of online sparse + low-rank matrix recovery from their sum. If the primary goal is to recover the low-dimensional subspace in which the l_t's lie, then the problem is one of online robust principal components analysis (PCA). An example of where such a problem might arise is in separating a sparse foreground and a slowly changing dense background in a surveillance video. In our work, we have developed a novel algorithm called ReProCS to solve this problem and demonstrated its significant advantage over other robust PCA based methods for the video layering problem. While there has been a large amount of recent work on performance guarantees for the batch robust PCA or the batch sparse + low-rank matrix recovery problem, the online problem is largely open. In recent work, we have shown that, with ReProCS, under mild assumptions and with high probability, the error in recovering the subspace in which l_t lies decays to a small value within a short delay of a subspace change time and the support of x_t is recovered exactly. Moreover, the error made in estimating x_t and l_t is small at all times. The assumptions that we need are (a) a good estimate of the initial subspace is available (easy to obtain using a short sequence of background-only frames in video surveillance); (b) the l_t's obey a `slow subspace change' assumption; (c) the basis vectors for the subspace from which l_t is generated are dense (non-sparse); and (d) the support of x_t changes by at least a certain amount at least every so often. (based on joint work with Cheniu Qiu and Brian Lois) BIO: Namrata Vaswani received a B.Tech. from the Indian Institute of Technology (IIT), Delhi, in 1999 and a Ph.D. from the University of Maryland, College Park, in 2004, both in Electrical Engineering. During 2004-05, she was a research scientist at Georgia Tech. Since Fall 2005, she has been with the Iowa State University where she is currently an Associate Professor of Electrical and Computer Engineering. She has held the Harpole-Pentair Assistant Professorship at Iowa State during 2008-09. From October 2009 to February 2013, she served as an Associate Editor for the IEEE Transactions on Signal Processing (TSP). She is the recipient of the 2014 Iowa State Early Career Engineering Faculty Research Award and the 2014 IEEE Signal Processing Society (SPS) Best Paper Award for her 2010 paper in TSP (jointly with her former graduate student Wei Lu). Vaswani's research interests lie at the intersection of signal and information processing and machine learning for high dimensional problems. She also works on applications in big-data, video analytics and bio-imaging. In the last several years her work has focused on developing provably accurate online algorithms for high-dimensional structured data recovery problems such as online sparse matrix recovery (recursive recovery of sparse vector sequences) or dynamic compressed sensing, online robust principal components' analysis (PCA) and online matrix completion; and on demonstrating their usefulness in dynamic magnetic resonance imaging (MRI) and video analytics.
March
3/05
4:30PM
Dr. Nicolas Christin
Carnegie Mellon University
Measuring and Defending Against Search-Result Poisoning
3/12
4:30PM
Dr. Abhishek Bhattacharjee
Rutgers University
ABSTRACT: Since its inception, virtual memory has become a powerful and ubiquitous abstraction for allocating and managing memory with a flexible and clean programming model. Typically, the systems community has been comfortable paying a performance tax for these programmability benefits. Unfortunately, emerging software with large data requirements and deeper stacks (e.g., large graphs, key value stores, virtualization), and emerging hardware accelerators requiring manual data orchestration by the CPU are increasing this performance tax drastically, while also conceding various programmability benefits of virtual memory. In this talk, I discuss techniques to reclaim this lost performance and programmability by enriching existing address translation hardware to more elasticity adapt to memory allocation aspects of the operating system. Specifically, I show how hardware support that detects patterns in page table allocation can be used to design low-overhead, high performance address translation hardware. In addition, I discuss how to design memory management units for accelerators in support of unified address spaces. Overall, these techniques are broadly applicable across both server and client systems. BIO: Abhishek Bhattacharjee is an assistant professor in the department of computer science at Rutgers University. His interests span the the interactions between architecture and operating systems. Abhishek received his PhD from Princeton University in 2010 and the NSF Career award in 2013.
3/19
4:30PM
Dr. Lee Potter
Ohio State University
ABSTRACT: Phase-contrast magnetic resonance imaging (PC-MRI) is a noninvasive tool to assess cardio-vascular disease by quantifying blood flow; however, long acquisition times presently limit the spatial and temporal resolutions, real-time imaging, and extensions to 4D flow imaging in clinical settings. We propose a novel technique for accelerated PC-MRI. The technique is based on Bayesian inference yet admits fast computation via an approximate message passing algorithm. The Bayesian formulation allows us to model and exploit the statistical relationships across space, time, and encodings in order to achieve reproducible estimation of flow from highly under-sampled data. The imaging approach combines physical modeling, experiment design, and principled, yet tractable, computation. In vivo prospectively down-sampled results are presented from five healthy volunteers imaged using a 1.5T Siemens scanner. We conclude with discussion of the proposed approach as an embodiment of generic principles for computational imaging, with a specific example from radar imaging. BIO: Lee C. Potter received the B.E. degree from Vanderbilt University, Nashville, TN, and the M.S. and Ph.D. degrees from the University of Illinois at Urbana-Champaign, all in electrical engineering. Since 1991, he has been with the Department of Electrical and Computer Engineering, The Ohio State University, Columbus, where he is also an investigator at the Davis Heart and Lung Research Institute. His research interests include statistical signal processing, inverse problems, detection, and estimation, with applications in radar and medical imaging. Dr. Potter is a recipient of the OSU College of Engineering MacQuigg Award for Outstanding Teaching.
3/23
12:0AM
No graduate seminar week of March 26th
April
4/02
4:30PM
Dr. Jose Carmena
University of California - Berkeley
Closed-Loop Design Strategies for Neuroprosthetic Control
4/09
4:30PM
Dr. Frank Lewis
University of Texas Arlington Research Institute
Cooperative Synchronization In Renewable Energy Microgrid Generation
4/16
4:30PM
Dr. Mario Gerla
University of California Los Angeles
ABSTRACT: As vehicles will soon become network connected, new vehicle applications are emerging, from navigation safety to location aware content distribution, urban surveillance and intelligent transport. Autonomous vehicles stand out as important players, with plenty of sensors, memory and processing power. The richness of on board resources and the diversity of applications set the Vehicular ad Hoc Network (VANET) apart from conventional MANETs and introduce new challenges in the services they provide. A representative service scenario is urban sensing: vehicles monitor the environment, classify the events, e.g., license plates, chemical readings, and support search requests from peer vehicles, from Authorities and from external Agents. This notion of service suggests that the VANET can be viewed as a Mobile Computing Cloud (MCC) where vehicles interact and collaborate to sense the environment, process the data, V2V propagate the results and share resources while providing mobile services. This talk will revisit VANET applications and will propose a Vehicular Cloud platform to support this service model. BIO: Dr. Mario Gerla is a Professor in the Computer Science Dept at UCLA. He holds an Engineering degree from Politecnico di Milano, Italy and the Ph.D. degree from UCLA. At UCLA, he was part of the
4/23
4:30PM
Dr. Christopher Rozell
Georgie Institute of Technology
4/30
4:30PM
Dr. Guillermo Villanueva
Ecole Polytechnique Federale de Lausane
NEMS Oscillators – Main challenges and workarounds. Does it really make sense?
No seminars scheduled.
Date
Time
Lecturer
Affiliation
Topic
Resources
Top of page ↑
September
9/03
4:30PM
Dr. Harish Krishnaswamy
Columbia University
Rethinking The Functional Boundaries of Integrated Radio-Frequency Systems Enables New Wireless Communication Paradigms
9/17
4:30PM
Dr. Narayana Santhanam
University of Hawaii
Statistics of slow mixing Markov processes with applications to community detection
9/24
4:30PM
Dr. Jiang Hu
Texas A&M University
October
10/01
4:0PM
Dr. Tim brown
Carnegie Mellon University
Abstract: Modern radio networks are a mess. The good old days of homogeneous base stations arranged in regular grids are long gone. Tiny femto cells mix with big macro cells, Cellphones mix with WiFi. Base stations are placed wherever their needed. How do we make sense of this? Should we expect it to work? Will I have good good coverage and throughput? How does this compare to the good old days? Will the billions of sensors in the Internet of Things be the technology that finally makes it fall apart? This talk describes a comprehensive study of the down-link performance in these so-called heterogeneous networks. We consider a general model consisting of an arbitrary number of open-access and closed-access tier of base stations arranged according to independent homogeneous Poisson point processes. For such a system, analytical characterizations for the coverage probability and average rate at an arbitrary mobile-station, and average per-tier load are derived. The results also demonstrate the effectiveness and analytical tractability of the stochastic geometric approach to study such complex radio systems' performance. Bio: Timothy X Brown received his B.S. in physics from Pennsylvania State University and his Ph.D. in electrical engineering from California Institute of Technology. He has worked at both the Jet Propulsion Laboratory and Bell Communications Research. Since 1995 he has been at the University of Colorado at Boulder, most recently as Professor in Electrical, Computer, and Energy Engineering and Director of the Interdisciplinary Telecommunications Program. He is currently a Distinguished Service Professor at Carnegie Mellon University in EPP, ECE, and the graduate programs in Kigali, Rwanda. His research interests include wireless communication systems, network security, and machine learning. His recent research funding includes NSF, DOE, and industry. Projects include the role of mobility in network control of unmanned aircraft, denial of service vulnerabilities in wireless protocols, spectrum policy frameworks for cognitive radios, and stochastic geometry applied to wireless networks. He is a recipient of the NSF CAREER Award, and the GWEC Wireless Educator of the Year Award.
10/08
4:30PM
Dr. Yuejie Chi
Ohio State University
Sparse Inversion of Mixture and Bilinear Models via Convex Optimization
10/15
4:30PM
Dr. Donatello Materassi
Identifying the underlying structure of networks of dynamical systems via passive observations
10/22
4:30PM
Dr. Oliver Cossairt
Northwest University
Computational Imaging and Illumination for Three Dimensional Imaging: Research at the NU Comp Photo Lab
10/29
4:30PM
Dr. Sennur Ukulus
University of Maryland
Energy Harvesting and Energy Cooperation in Wireless Communications
November
11/05
4:30PM
Dr. Vijay Balasbramanian
University of Pennsylvania
11/12
4:30PM
Dr. Jonathan Pillow
Princeton
11/19
4:30PM
Dr. Marilyn Wolf
Georgia Institute of Technology
December
12/03
4:30PM
Dr. Alanson Sample
Disney Research
12/10
4:30PM
Dr. Chee-Wooi Ten
Michigan Technological University
Cyber-Based Contingency Analysis

Related seminar series at Carnegie Mellon

Other seminar series