Spring 2017 SIP Seminars
1/25/2017 - A User-Centric Resource Management and Optimization Framework for Multi-* Wireless Networks - Presented by Dr. Eirini Eleni Tsiropoulou (U. of Maryland)
Dr. Eirini Eleni Tsiropoulou
Title: A User-Centric Resource Management and Optimization Framework for Multi-* Wireless Networks
Abstract: The proliferation of emerging networks and architectures, wireless access technologies and multi-homing smart devices, create a competitive environment where users have access to various providers, resources and are offered with multiple service options. The latter is intensified with the emergence of several new paradigms including Internet of Things (IoT) and Device-to-Device (D2D) communications that increase both complexity and flexibility to unprecedented levels. This fact demands and motivates the development of user-centric resource management and optimization frameworks, which enable user’s self-optimization and autonomicity. In such a competitive and distributed environment game theory arises as a natural choice and a powerful theoretical tool to cope with the corresponding problems, while properly capturing and reflecting users’ behavior within the competitive arena of system’s resource allocation. In the main part of the talk, the formulation of resource allocation problems in wireless communication systems is discussed and game theoretic approaches are proposed to determine stable solutions (e.g. Nash equilibrium points). Multi-* wireless communication networks are examined covering and reflecting several dimensions of heterogeneity, diversity and multiplicity, such as multi-tier architectures (e.g. cellular, femtocells, VLC), multi-resources in nature (e.g., power and/or rate control) or in properties (e.g., continuous and/or discrete, depending on the wireless access technology), multi-services (e.g. reflecting different QoS requirements) and multiproviders. At the last part of the talk, insight information about applying the previously presented techniques in resource management problems in other emerging and relevant application domains, such as smart grid networks and Internet of Things environment will be discussed.
Biography: Eirini Eleni Tsiropoulou is a postdoctoral researcher at Institute for Systems Research, Department of Electrical and Computer Engineering, University of Maryland. She obtained her Diploma, MBA in techno-economics and Ph.D. degree in Electrical and Computer Engineering from the National Technical University of Athens in 2008, 2010 and 2014 respectively. Before joining UMD she was a Senior Research Associate at the National Technical University of Athens and a visiting scholar at the University of Texas at Dallas. Two of her papers received the Best Paper Award at IEEE Wireless Communications and Networking Conference (WCNC 2012) in April 2012 and at the 7th International Conference on Ad Hoc Networks (ADHOCNETS 2015) in September 2015. Her main research interests lie in the area of wireless heterogeneous networks, with emphasis on optimization and resource allocation, smart data pricing, Internet of Things, game theory, social recommendation and personalization for multimedia, and demand response management in smart grid networks. She has actively participated holding leading roles in several research and development proposals and projects funded by the European Commission, in the broader area of autonomic wireless networks, optimization in future internet architectures and multimedia content delivery systems.
2/8/2017 - How Busy Are You? Predicting the Interruptibility Intensity of Mobile Users - Presented by Prof. Janne Lindqvist (Rutgers)
Prof. Janne Lindqvist
Title: How Busy Are You? Predicting the Interruptibility Intensity of Mobile Users
Abstract: Smartphones frequently notify users about newly available messages or other notifications. It can be very disruptive when these notifications interrupt users while they are busy. Our work here is based on the observation that people usually exhibit different levels of busyness at different contexts. This means that classifying users’ interruptibility as a binary status, interruptible or not interruptible, is not sufficient to accurately measure their availability towards smartphone interruptions. In this paper, we propose, implement and evaluate a two-stage hierarchical model to predict people’s interruptibility intensity. Our work is the first to introduce personality traits into interruptibility prediction model, and we found that personality data improves the prediction significantly. Our model boot- straps the prediction with similar people’s data, and provides a good initial prediction for users whose individual models have not been trained on their own data yet. Overall prediction accuracy of our model can reach 66.1% while the first-stage binary interruptibility prediction accuracy is 75%. (This is a practice talk for CHI’17, and the paper is available here: http://www.winlab.rutgers.edu/~janne/CHI17-predictinginterruptibility.pdf)
Biography: Janne Lindqvist is an assistant professor of electrical and computer engineering and a radiant member of WINLAB at Rutgers University. His work is frequently featured in the popular media with close to thousand mentions so far including several times in IEEE Spectrum, MIT Technology Review, Scientific American, Communications of the ACM, NPR, WHYY Radio, Yahoo! News, International Business Times, and recently also in ABC News Radio, CBS Radio News, Fortune, Computerworld, Der Spiegel, London Times, Slashdot, The Register. Janne has two engineering graduate degrees in computer science and engineering. Most of his academic contributions to science and humanity are in security engineering, while he has heard local colleagues sometimes refer to him as a “sociologist” or “expert in human-computer interaction.” To be fair to his colleagues, Janne indeed directs the Rutgers Human-Computer Interaction group. Janne’s work focuses on hard real-world problems, and currently his group and his colleagues work includes usable and secure authentication, mobile privacy, physical-world crowdsourcing, measuring implicit racism in situ, social protocols for wireless networking, and ecological field studies on non-suicidal self-injurous behavior. He is awards include the Best Paper Award from MobiCom’12, the Best Paper Nominee Award from UbiComp’14, Sustainable Jersey Creation & Innovation Award 2014.
2/22/2017 - Human Action Attribute Learning From Video Data Using Low-Rank Representations - Presented by Tong Wu (Rutgers)
Title: Human Action Attribute Learning From Video Data Using Low-Rank Representations
Abstract: Representation of human actions as a sequence of human body movements or action attributes enables the development of models for human activity recognition and summarization. In this talk, I will present an extension of the low-rank representation (LRR) model, termed the clustering-aware structure-constrained low-rank representation (CS-LRR) model, for unsupervised learning of human action attributes from video data. This model is based on the union-of-subspaces (UoS) framework, and integrates spectral clustering into the LRR optimization problem for better subspace clustering results. I will also introduce a hierarchical subspace clustering approach, termed hierarchical CS-LRR, to learn the attributes without the need for a priori specification of their number. By visualizing and labeling these action attributes, the hierarchical model can be used to semantically summarize long video sequences of human actions at multiple resolutions. A human action or activity can also be uniquely represented as a sequence of transitions from one action attribute to another, which can then be used for human action recognition.
Biography: Tong Wu received BE degree in Instrument Science and Engineering from Shanghai Jiao Tong University, China in 2009, and MS degree in Electrical Engineering from Duke University in 2011. Since September 2012, he has been working towards the PhD degree at the Department of Electrical and Computer Engineering, Rutgers, The State University of New Jersey. He is the recipient of the Best Student Paper Award at the 12th IEEE Image Video and Multidimensional Signal Processing (IVMSP) Workshop. His research interests include big data analytics, machine learning, image and video processing, and statistical signal processing.
3/1/2017 - Efficient Redundancy Techniques to Reduce Delay in Cloud Systems - Presented by Prof. Gauri Joshi (CMU)
Prof. Gauri Joshi
Title: Efficient Redundancy Techniques to Reduce Delay in Cloud Systems
Abstract: Ensuring fast and seamless service to users is critical for today’s cloud services. However, guaranteeing fast response can be challenging due to random service delays that are common in today’s data centers. In this talk I explore the use the redundancy to combat such service variability. For example, replicating a computing task at multiple servers and then waiting for the earliest copy saves service time. But the redundant tasks can cost more computing resources and also delay subsequent tasks. I present a queueing-theoretic framework to answer fundamental questions such as: 1) How many replicas to launch? 2) Which queues to join? 3) When to issue and cancel the replicas? This framework reveals surprising regimes where replication reduces both delay as well as resource cost. The task replication idea can also be generalized to analyze latency in content download from erasure coded storage. More broadly, this work lays the theoretical foundation for studying queues with redundancy, uncovering many interesting future directions in cloud infrastructure, crowdsourcing and beyond.
Biography: Bio: Gauri Joshi is a Research Staff Member at IBM T. J. Watson Research Center. She will be joining the Carnegie Mellon ECE department as assistant professor in Fall 2017. Gauri completed a Ph.D from MIT EECS in June 2016. Before coming to MIT, she completed a B.Tech and M. Tech in Electrical Engineering from the Indian Institute of Technology (IIT) Bombay in 2010. Her awards and honors include the Best Thesis Prize in Computer science at MIT (2012), Institute Gold Medal of IIT Bombay (2010), Claude Shannon Research Assistantship (2015-16), and Schlumberger Faculty for the Future fellowship (2011-2015).
Title: Robust Estimator for Multiple Inlier Structures
Abstract: Robust regression separates inliers from outliers and estimates the objective function of the inlier structures. A new robust estimator is presented which does not require any user given parameters, except the number of random sampling. It detects each structure independently and can handle different inlier scales of these structures. After all the input data was classified, the structures are sorted by their strengths and the strongest inlier structures come out at the top. Like any robust estimator, this algorithm also has limitations, which will be described in details. The efficiency and accuracy of this method will be demonstrated by several 2D computer vision problems and 3D point cloud applications.
Biography: Xiang Yang received the BE degree in mechanical engineering and automation from Beihang University, Beijing, China, in 2009, and the MS degree in mechanical engineering from University of Bridgeport, Connecticut, in 2012. Currently, he is working toward the PhD degree in mechanical and aerospace engineering at Rutgers University, New Jersey. His research interests include computer aided design, 3D reconstruction and statistical pattern recognition.
3/29/2017 - An Information Theoretic Perspective of Consistent Distributed Storage - Presented by Prof. Viveck Cadambe (Penn State)
Prof. Viveck Cadambe
Title: An Information Theoretic Perspective of Consistent Distributed Storage
Abstract: Distributed key-value stores implementations are an integral part of modern cloud computing infrastructure, and are used by various applications including transactions, reservation systems, multi-player gaming and multi-processor programming. The principles of modern key-value store design are closely tied to a problem called consistent shared memory emulation, which is studied in distributed computing theory. The goal of consistent shared memory emulation is to implement a read-write data object in a distributed storage system. In shared memory emulation, it is important to be resilient to server failures, to allow concurrent access to external clients, and to ensure the following property known as consistency: when the data is being constantly updated, a client that reads from the system should obtain the latest (consistent) version of the data. Motivated by technological trends where key-value stores are increasingly implemented in high speed memory, we will use develop and use information theory ideas to understand and minimize the memory footprint (storage overhead) of consistent shared memory emulation. In this talk, we present an overview of three main ideas. First, we present an atomically consistent shared memory algorithm that uses classical erasure codes in its storage scheme. The algorithm exposes the salient challenges of using erasure coding in shared memory emulation. Second, we present a new, relatively simplified, information theory framework that enables us to study the memory footprint of consistent shared memory emulation. Our framework opens the door to the development and use of compression and coding techniques to minimize the memory footprint of shared memory emulation. Third, time-permitting, we will show connections between our simplified information-theoretic formulation and the full-fledged shared memory emulation model via a new impossibility result.
Biography: Viveck Cadambe is an Assistant Professor in the Department of Electrical Engineering at Pennsylvania State University. Dr. Cadambe received his Ph.D from the University of California, Irvine in 2011. Between 2011 and 2014, he was a postdoctoral researcher, jointly with the Electrical and Computer Engineering (ECE) department at Boston University, and the Research Laboratory of Electronics (RLE) at the Massachusetts Institute of Technology (MIT). Dr. Cadambe is a recipient of the 2009 IEEE Information Theory Society Best Paper Award, the 2014 IEEE International Symposium on Network Computing and Applications (NCA) Best Paper Award, the 2016 NSF Career Award and a finalist for the 2016 Bell Labs Prize. His research involves understanding of data communication and storage using the tools of information theory and coding theory. His interests include applications to wireless communication networks, and cloud storage and computing systems.
Title: On Sure Independence Screening
Abstract: With the deluge of high-dimensional data and advances in computational power, variable selection and screening have become important concepts in many scientific studies. In this work, we study the concept of sure screening in the context of screening using marginal correlations. The main result of our work shows that a condition, which we term as the screening condition, is sufficient for the sure screening property to hold for the simple marginal correlations based screening procedure. The practicality of the screening condition is demonstrated by evaluating the screening condition for sub-Gaussian as well as arbitrary (random or deterministic) matrices to provide sufficient conditions for the sure screening of such matrices. Furthermore, the theoretical guarantees for sure screening of sub-Gaussian and arbitrary matrices are compared to known results for sure screening. Our theoretical guarantees also offer new insights on the marginal correlations based screening procedure, and our numerical results corroborate these insights.
Biography: Talal Ahmed is an ECE PhD student working under the supervision of Prof. Waheed Bajwa at Rutgers University. He received his BS at LUMS, Pakistan in 2012 and his MS at Rutgers University in 2016. His research interests include theories and methods in high-dimensional statistics.
4/26/2017 - A Robust Sparse Fourier Transform and Its Application in Radar Signal Processing - Presented by Shaogang Wang (Rutgers)
Title: A Robust Sparse Fourier Transform and Its Application in Radar Signal Processing
Abstract: We propose the Robust Sparse Fourier Transform (RSFT), a tool that enables the application of the Sparse Fourier Transform (SFT) to a real world, noisy setting. The RSFT can accommodate off-grid frequencies in the data. Furthermore, by incorporating Neyman-Pearson detection in the SFT stages, frequency detection in the RSFT does not require knowledge of the exact sparsity of the signal and is robust to noise. We analyze the asymptotic performance of the RSFT, and study the computational complexity versus detection performance trade-off. We show that by appropriate choice of the detection thresholds, the optimal trade-off can be achieved. We discuss the application of RSFT on short range ubiquitous radar signal processing, and demonstrate its feasibility via simulations.
Biography: Shaogang Wang is a Ph.D. candidate in the ECE department of Rutgers University, working with Prof. Athina Petropulu and Prof. Vishal Patel. His research interest is in Sparse Signal Processing, with an emphasis on radar applications in a real-life scenario. Shaogang received the B.E. degree in Electrical Engineering from Nanjing University of Aeronautics and Astronautics, Nanjing, China, in 2004, and the M.S. degree in Signal and Information Processing from Shanghai Academy of Spaceflight Technology, Shanghai, China, in 2007. He was with Shanghai Institute of Spaceflight Electronics Technology, Shanghai, China, from March 2007 to July 2014, focusing on radar signal processing and system design.