Seminars: Spring 2015
Dimensionality reduction techniques for video processing and retrieval
Speaker:Anthony Vetro, Mitsubishi Electric Research Labs, Cambridge, MA
Time: 11:00 am - 12:00 pm, Jan 29, 2015
Location: 2MTC, 9th floor, Room 9.101, Brooklyn, NY
This talk will describe several techniques for dimensionality reduction and their application for video processing and retrieval. First, a factorized robust matrix completion algorithm is described to solve the video background subtraction problem. The algorithm decomposes a sequence of video frames into the sum of a low rank background component and a sparse motion component. The next part of the talk discusses how similar low-rank matrix factorization techniques can be used to realize a compact visual feature representation for content retrieval. Finally, an approach based on quantized embeddings is presented as means to achieve a rate-efficient representation of high-dimensional features such that pair-wise distances between the underlying feature vectors are preserved.
About the Speaker: Anthony Vetro is a Deputy Director at Mitsubishi Electric Research Labs, Cambridge, MA, and also manages a group that is responsible for research on video coding and image processing, information security, sensing technologies, and speech/audio processing. He received the B.S., M.S., and Ph.D. degrees in electrical engineering from Polytechnic University, Brooklyn, NY. He has published more than 200 papers, and has been an active member of ISO/IEC and ITU-T standardization committees on video coding for many years. Dr. Vetro is also active in various IEEE conferences, technical committees, and editorial boards, and is a Fellow of IEEE. Web site: http://www.merl.com/people/avetro
A Framework for Structural Input/Output and Control Configuration Selection of Large - Scale Systems
Speaker:Sérgio Pequito, University of Pennsylvania
Time: 11:00 am - 12:00 pm, Jan 30, 2015
Location: LC433, 5 MetroTech Center, Brooklyn, NY
The structure control system design consists mainly of two steps: input/output (I/O) selection and control configuration (CC) selection. The first one is devoted to the problem of computing how many actuators/sensors are needed and where should be placed in the plant to obtain some desired property. Control configuration is related to the decentralized control problem and is dedicated to the task of selecting which outputs (sensors) should be available for feedback and to which inputs (actuators) in order to achieve a predefined goal. The choice of inputs and outputs affects the performance, complexity and costs of the control system. Due to the combinatorial nature of the selection problem, an efficient and systematic method is required to complement the designer intuition, experience and physical insight. Motivated by the above, this presentation addresses the structure control system design taking explicitly into consideration the possible application to large - scale systems. We provide an efficient framework to solve the following major minimization problems: i) selection of the minimum number of manipulated/measured variables to achieve structural controllability/observability of the system, and ii) selection of the minimum number of measured and manipulated variables, and feedback interconnections between them such that the system has no structural fixed modes. Contrary to what would be expected, we showed that it is possible to obtain the global solution of the aforementioned minimization problems in polynomial complexity in the number of the state variables of the system. To this effect, we propose a methodology that is efficient (polynomial complexity) and unified in the sense that it solves simultaneously the I/O and the CC selection problems. This is done by exploiting the implications of the I/O selection in the solution to the CC problem.
About the Speaker: Sérgio Pequito is a postdoctoral researcher at University of Pennsylvania. He obtained his PhD in Electrical and Computer Engineering from Carnegie Mellon University and Instituto Superior Técnico, through the CMU-Portugal program. Furthermore, he received his BSc and MSc in Applied Mathematics from the Instituto Superior Técnico. Pequito's research consists in understanding the global qualitative behavior of large scale systems from their structural or parametric descriptions and provide a rigorous framework for the design, analysis, optimization and control of large scale (real-world) systems. Pequito was awarded with the best student paper finalist in the Conference on Decision and Control 2009, the ECE Outstanding Teaching Assistant Award at Carnegie Mellon University, and the Carnegie Mellon Graduate Teaching Award (university-wide) honorable mention, both in 2012.
Capacity management in stochastic loss network models
Speaker: Yingdong Lu, IBM T.J. Watson Research Center
Time: 11:00 am - 12:00 pm, Feb 5, 2015
Location: 2MTC, 10th floor, Room 10.099, Brooklyn, NY
Motivated by capacity management problems in cloud computing and business analytics, we study a set of stochastic optimization problems centered around the analysis of stochastic loss networks. Several different methods of calculating loss probability, a key performance metric, are developed for different applications. We will discuss both the theoretical development and numerical experiments for these methods.
About the Speaker: Dr. Yingdong Lu is a research staff member in the department of mathematical sciences at IBM T.J. Watson Research Center. He received a B.S. degree in Mathematics from Peking University, a Ph.D in Operations Research from Columbia University. His main interests include stochastic processes, queueing and queueing network, and inventory and supply chain management.
From biomedical imaging to online blogs: Graph signal processing
Speaker:Jelena Kovačević, Carnegie Mellon University
Time: 11:00 am - 12:00 pm, Feb 12, 2015
Location: 2MTC, 10th floor, Room 10.099, Brooklyn, NY
I will present a path from classification in biomedical imaging to online blogs, where a common thread is graph signal processing, a theoretical framework that generalizes fundamental concepts of classical signal processing from regular domains, such as lines and rectangular lattices, to general graphs. It is particularly applicable to domains such as physical, engineering, and social, where signals are characterized by irregular structure. Signal processing on graphs has found multiple applications, including approximation, sampling, classification, inpainting and clustering, and I will describe some of these.
About the Speaker: Jelena Kovačević received a Ph.D. degree from Columbia University. She then joined Bell Labs, followed by Carnegie Mellon University in 2003, where she is currently the Edward David Schramm Professor and Head of the Department of ECE, and Professor of BME. She received the Dowd Fellowship at CMU, Belgrade October Prize, and the E.I. Jury Award at Columbia University. She is a coauthor on an SP Society award-winning paper and is a coauthor of the textbooks Wavelets and Subband Coding and Foundations of Signal Processing. Dr. Kovacevic is the Fellow of the IEEE and was the Editor-in-Chief of the IEEE Transactions on Image Processing. She was a keynote speaker at a number of meetings and has been involved in organizing numerous conferences. Her research interests include multiresolution techniques, graphs, biomedical imaging, and smart infrastructure.
Information Security in Controlled Dynamical Systems: Trading Utility for Privacy
Speaker: Parv Venkitasubramaniam, Lehigh University
Time: 11:00 am - 12:00 pm, Feb 13, 2015
Location: LC433, 5 MetroTech Center, Brooklyn, NY
Cyber physical systems, which rely on the joint functioning of information and physical systems, are vulnerable to information leakage through the actions of the internal physical control system. In particular, if an external observer has access to input and output variables of the system exposed through cyber communication links, then critical information can be inferred about the internal states of the system and consequently compromise the security of system operation. In this talk, a mathematical framework based on Markov Decision Process models will be discussed to investigate the design of controller actions when an internal state privacy requirement is imposed as part of the system objective. Fundamental tradeoffs will be presented between the privacy achievable-- as measured using information theoretic equivocation-- and the tangible utility of the internal controller. Two examples will be discussed that apply the developed model-- the first maximizes user anonymity in data networks under memory and fairness restrictions, and the second maximizes privacy of energy usage from smart metered electricity measurement using an in-home energy storage mechanism.
About the Speaker: Parv Venkitasubramaniam is an assistant professor in the Electrical and Computer Engineering department at Lehigh University. His current research interests are in developing theoretical foundations for privacy and security for networked cyber and cyber physical systems. He received a B. Tech degree from the Indian Institute of Technology, Madras and graduated with a Ph.D from Cornell University under the advise of Prof. Lang Tong. Prior to joining Lehigh he was a visiting post-doctoral researcher at University of California, Berkeley hosted by Prof. Venkat Anantharam. Parv is a recipient of the 2005 IEEE Leonard G. Abraham Award and a 2012 NSF CAREER Award.
Mind the Map! The Impact of Subway Map on Passengers' Route Choice Decisions in London and Washington DC
Speaker:Zhan Guo, New York University
Time: 11:00 am - 12:00 pm, Feb 26, 2015
Location: 2MTC, 10th floor, Room 10.099, Brooklyn, NY
Subway users often rely on schematic system maps to identify the best route from A to B. However, the schematic map is always distorted. Some routes may appear longer or shorter on the map than in reality. The line alignment also differs between a schematic map and an actual map. Such distortion may “misguide” passengers to favor a particular route over another, thus “rebalance” the passenger flows inside a subway network, creating or relieving bottlenecks in the system. If this is true, schematic subway maps could become an effective planning tool to influence route decisions for the benefits of not only the system but also individual passengers.
The London study, based on actual route choices in the London Underground, tests the first assumption, and confirms that when choosing a route most passengers would trust the schematic map more than their own experience of reality. This is true even for the most experienced passengers who have frequently used the system for many years. The Washington DC study, based on an experiment in the crowdsourcing platform--Amazon’s Mechanic Turk, tests the second assumption, and illustrates that changing the length and alignment of lines on a schematic map can switch passengers from bottlenecks to underutilized segments in the DC subway, thus improving the efficiency of the system. Depending on how the lines are exactly modified, the “switching” effect also differs.
About the Speaker: Zhan Guo studies individuals’ travel behavior and explores innovative ways to influence the decision-making process to produce better social outcomes such as reduced congestion and carbon emissions. At the micro level, he focuses how travelers perceive travel alternatives and attributes and what discrepancies exist between perception and reality. The ability to reinforce, change, or even deceive that perception to promote the "right" behavior, and the methods used to do so, also figure largely in his research. At the macro level, he is interested in the effect of technical standards, such as parking and street standards, on the built environment and the rationale behind these standards. The (dis)connection between government regulations, market forces, and consumer preferences is the focus of his research. Zhan has conducted empirical work on transfer behavior in Boston, metro map design in London, parking policy in New York, and pedestrian environments in Hong Kong. His work has been covered by New York Times, Wall Street Journal, BBC, Economist, Le Monde, ABC Evening News, the Atlantic Cities, Nudges.org, etc.
Zhan Guo received a B. Arch from Tianjin University, a MUD from Tsinghua University, China, and a MCP and a Ph.D in Urban Planning from MIT.
Proximal Splitting Algorithms in Data Sciences
Speaker:Patrick L. Combettes, Pierre and Marie Curie University, Paris
Time: 11:00 am - 12:00 pm, Mar 2, 2015
Location: LC433, 5 MetroTech Center, Brooklyn, NY
In recent years proximal splitting algorithms have become prominent tools for solving a wide array of data-driven problems. We provide an overview of the field and report on recent developments in the area of complex structured optimization problems. In particular, we present a new block-coordinate approach to tackle problems of very large sizes. Applications to machine learning, inverse problems, and image recovery will be discussed.
About the Speaker: P. L. Combettes received his Ph.D. degree from North Carolina State University in 1989. In 1990, he joined the faculty of the City College and the Graduate Center of the City University of New York, where he became a Full Professor in 1999. Since 2000, he has been with the Faculty of Université Pierre et Marie Curie - Paris 6, laboratoire Jacques-Louis Lions, where he is presently a Professeur de Classe Exceptionnelle. He was elected a Fellow of the IEEE in 2005.
Systems, Algorithms, and Applications for Robotic Sensing
Speaker:Pratap Tokekar, University of Pennsylvania
Time: 11:00 am - 12:00 pm, Mar 5, 2015
Location: 2MTC, 10th floor, Room 9.009, Brooklyn, NY
A connected network of robots, sensors, and smart devices has the potential to solve grand challenges in domains such as agronomy, oceanography, and emergency response. Robots will form the "physical" layer of this network and collect data from hard to reach places at unprecedented spatio-temporal scales. A key challenge for roboticists is to design decision-making algorithms that enable optimal sensing and data collection. In this talk, I will present such algorithms for three types of problems: (i) How to cover complex environments using few sensors and robots? (ii) How to optimize the robots' energy to enable long term operation? (iii) How to coordinate teams of robots and sensors?
In the first part, we will focus on the algorithmic aspects of the Art Gallery Problem, a classical sensor placement problem. I will present a new formulation that guarantees a good view of convex objects in polygonal environments, despite self-occlusions. I will present bounds and approximation algorithms for placing minimum number of sensors for this new formulation. I will also present algorithms for robotic coverage and informative path planning problems. In the second part, I will describe techniques for efficient solar energy harvesting and planning energy-aware trajectories for collaborative aerial and ground robots. In the third part, I will present coordination algorithms for target tracking with robotic teams that respects the communication constraints between the robots.
We will study these problems in the context of two practical applications: precision agriculture and autonomous monitoring radio-tagged fish in Minnesota lakes. Throughout the talk, along with theoretical results, I will present experiments conducted with autonomous boats in Minnesota lakes, wheeled robots operating on frozen lakes, and aerial robots operating in corn plots and apple orchards.
About the Speaker: Pratap Tokekar is a Postdoctoral Researcher at the University of Pennsylvania. He obtained his Ph.D. in Computer Science from the University of Minnesota in 2014 and Bachelor of Technology degree in Electronics and Telecommunication from College of Engineering Pune, India in 2008. His research interests include algorithmic and field robotics, sensor networks, and computational geometry.
Sparse Modeling for High-Dimensional Multi-Manifold Data Analysis
Speaker:Ehsan Elhamifar, University of California, Berkeley
Time: 11:00 am - 12:00 pm, Mar 9, 2015
Location: 2MTC, 10th floor, Room 10.099, Brooklyn, NY
One of the most fundamental challenges facing scientists and engineers across different fields, such as signal/image processing, computer vision, robotics and bioinformatics, is the large amounts of high-dimensional data that need to be analyzed and understood. In this talk, I present provably correct and efficient algorithms, based on the sparse representation theory, for the analysis of high-dimensional datasets by exploiting their underlying low-dimensional structures. I talk about algorithms for the two fundamental problems of clustering and subset selection in unions of subspaces and discuss the robustness of the algorithms to data nuisances. I show that these tools effectively advance the state-of-the-art data analysis in a wide range of important real-world problems, such as segmentation of motions in videos, clustering of images of objects, energy disaggregation and identification of hybrid dynamical systems.
About the Speaker: Ehsan Elhamifar is a postdoctoral scholar in the department of Electrical Engineering and Computer Sciences at the University of California, Berkeley. He obtained his PhD in Electrical and Computer Engineering from the Johns Hopkins University. Ehsan is broadly interested in developing provably correct and efficient data analysis algorithms that can address challenges of complex and large-scale high-dimensional datasets. Specifically, he focuses on the intrinsic low-dimensionality of real data and uses tools from convex geometry and analysis, sparse / low-rank representation, high-dimensional statistics and graph theory to develop such algorithms. Ehsan obtained MS and MSE degrees in Electrical Engineering and Applied Mathematics and Statistics, respectively, from Sharif University of Technology in Iran and the Johns Hopkins University.
Boosting Multimedia Content Analysis with Machine Learning Techniques
Speaker:Zhigang Ma, Carnegie Mellon University
Time: 11:00 am - 12:00 pm, Mar 11, 2015
Location: 2MTC, 10th floor, Room 10.099, Brooklyn, NY
Image and video analysis is the fundamental problem in multimedia community. In this talk, I will present how we improved the performance of multimedia content analysis by leveraging the advance of machine learning. The primary techniques harnessed in our work are comprised of feature selection, semi-supervised learning, learning an intermediate representation and knowledge adaptation. I will start from image annotation. In the literature, many different types of feature have been proposed to capture the semantic information of images. Impressive progress on image analysis has been witnessed based on these feature representations. However, it is inevitable that the feature representation has certain amount of noise and redundancy. Consequently, I will introduce a feature selection algorithm that is able to get a more compact representation, thus improving the annotation accuracy.
We also noticed that multimedia analysis faces a reality that precisely labeled images and videos are difficult to obtain, which is detrimental to learning a robust classifier. To address this issue, we proposed a semi-supervised learning framework for image and video analysis. Our approach is based on semi-supervised learning and it simultaneously considers eliminating feature noise and redundancy. Through extensive experiments on image and video classification, we validate that properly utilizing unlabeled data does contribute to the performance boost.
To step further, I will talk about a more challenging task in multimedia community, i.e., multimedia event detection. A multimedia event is a higher-level semantic abstraction of video sequences than a concept and consists of multiple concepts. As a multimedia event builds upon several basic elements of objects, scenes and human actions we propose to learn an intermediate representation coupled with the classifier learning by exploiting external resources. Our method is capable of learning an optimal event detector that carries more informative cues from the intermediate representation.
Lastly, I will introduce how to exploit knowledge adaptation to improve the performance of multimedia event detection when only few positive exemplars are given. The method is based on the assumption that multimedia events consist of low-level concepts and we proposed to adapt the knowledge from concept level to assist in event detection. Specifically, we use the available video corpora with annotated concepts as our auxiliary resource. Our approach has another desirable property that it is able to adapt knowledge from the source to the target even if the features of them are partially different, but overlapping.
About the Speaker: Zhigang Ma received the Ph.D. in computer science from University of Trento, Trento, Italy, in 2013. He is now a Postdoctoral Research Fellow with the School of Computer Science, Carnegie Mellon University, Pittsburgh, PA. His research interest is mainly on machine learning and its applications to multimedia analysis and computer vision. He has authored or co-authored more than 20 scientific articles at top venues, including the IEEE T-PAMI, T-MM, IJCV, ACM MM, CVPR, AAAI and IJCAI. He was a PC member for ACM MM 2014; a TPC member for ICME 2014 and 2015; a TPC member for ICMR 2015 and a PC member for IJCAI 2015. He is also an invited reviewer for IEEE Transactions on Multimedia, IEEE Transactions on Cybernetics, Multimedia Tools and Applications, Neurocomputing, Computer Vision and Image Understanding. Dr. Ma received the Outstanding PhD thesis award from SIGMM and the best PhD thesis award from Gruppo Italiano Ricercatori in Pattern Recognition, Italy.
Aerial autonomy at insect scale: What flying insects can tell us about robotics and vice versa
Speaker:Sawyer Buckminster Fuller, Harvard University
Time: 11:00 am - 12:00 pm, Mar 12, 2015
Location: 2MTC, 9th floor, Room 9.009, Brooklyn, NY
Insect-sized aerial robots will be deployed where their small size, low cost, and maneuverability give them an advantage over larger robots. For example, they could follow airborne plumes to locate methane leaks in dense piping infrastructure. Or deploy in swarms to perform detailed environmental monitoring. However, miniaturization poses challenges because scaling physics dictates that many conventional approaches used in larger aircraft, from electric motors and fixed wings, to the Global Positioning System and general-purpose microprocessors, cannot operate effectively at the size of insects. Insects have overcome these challenges, evolving a superlatively robust and agile flight apparatus that can land on flowers buffeted by wind or deftly avoiding a flyswatter. Though a detailed understanding of their flight systems lacking, they inspire robots with similar performance. My research is aimed at understanding the constraints inherent to autonomy -- that is, performing without a human operator -- at the insect scale. This includes investigations into the flight systems of flies and forward engineering robotic equivalents. I will describe experiments that revealed how flies compensate for wind disturbances in flight, and flight tests of insect-inspired controllers for fly-sized aerial robots. The results indicate that the severe power and weight constraints at this scale will require designs in which mechanisms, sensing, and computation are intricately interconnected. This suggests a multi-disciplinary design process that is reminiscent of evolution.
About the Speaker: Sawyer Buckminster Fuller is currently a Postdoctoral Scholar in the Microrobotics Laboratory of Prof. Robert Wood at Harvard. He creates biologically-inspired sensors, control systems, and mechanical designs targeted at insect-sized air vehicles and investigates the flight systems of aerial insects. He completed his Ph.D. in Biological Engineering at the California Institute of Technology and B.S. and M.S. degrees in Mechanical Engineering at the Massachusetts Institute of Technology. In addition to his work in insect flight control, he has also developed a frog-hopping robot at the NASA Jet Propulsion Laboratory and invented a 3D ink-jet microfabrication technology at the MIT Media Lab. His work at the intersection of robotics and biology has appeared in Science and Proceedings of the National Academy of Sciences, and been covered in news outlets such as Wired and The Scientist, and in an article on the cover of The MIT Technology Review.
Scenario-Based Model Predictive Control
Speaker:Georg Schildbach, University of California Berkeley
Time: 11:00 am - 12:00 pm, Mar 13, 2015
Location: LC433, 5 MetroTech Center, Brooklyn, NY
Model Predictive Control (MPC) is a powerful approach for optimal control of multi-variable systems with constraints. Its concept is to repeatedly solve a Constrained Finite-Horizon Optimal Control Problem (CFHOCP) on-line by numerical optimization. The CFHOCP requires predictions of the controlled system and the environment to be made. In reality, these predictions are subject to uncertainty, resulting from model errors, disturbances, and the uncertain evolution of the environment. In this context, both robust optimization (Robust MPC) and stochastic optimization (Stochastic MPC) have been well explored for handling this uncertainty. However, both approaches suffer from a variety of potential drawbacks, as discussed in this talk. Instead, we explore a much less treaded path that has proven to be very successful in estimation: particles, or in the context of MPC: scenarios. We will discuss the practical aspects of Scenario-Based MPC by looking at two different examples. We will also show that decision making based on scenarios leads into a deep mathematical theory. Recently, great advances were made in this field, but many open questions remain.
About the Speaker: Dr. Georg Schildbach obtained his Masters degrees in Applied Mechanics (Dipl.-Ing.) and Industrial Engineering (Dipl. Wirtsch.-Ing.), both at the Technical University of Darmstadt in Germany. He worked in various positions in the field of quantitative finance and investment banking for two years. Then he joined the Automatic Control Laboratory at ETH Zurich in 2010, where obtained his Ph.D. degree in the field of control and optimization in 2014. He currently holds the position of an Associate Director at the Hyundai of Excellence at UC Berkeley. His research interests evolve around algorithms for optimal and constrained control in uncertain environments, and their application to real-life problems in engineering and finance. His current research focus lies on developing new control algorithms for autonomous driving and intelligent transportation systems.
Semantic processing of visual signals in the era of data abundance
Speaker:Ruonan Li, Harvard University
Time: 11:00 am - 12:00 pm, Mar 23, 2015
Location: 2MTC, 10th floor, Room 10.009, Brooklyn, NY
What does the era of unprecedented amount of images and videos mean to semantic signal processing and machine learning? For inferring simple semantics such as object categories, identities, and event types, two predominant challenges are annotation poverty combined with data-selection bias, and the increased variations among the instances of the same semantic concept. Beyond simple semantics, richer semantics are emerging, such as social interactions and social relationships, to which current machines remain quite blind. I present three research vignettes that account for these challenges and opportunities. The first one introduces one of the earliest unsupervised domain adaption mechanisms that transfers knowledge from `in-house' training data (source domain) to a target domain, by exploiting geodesics in the space of factor analyzers, and I will also show its application to natural language processing. The second one develops a Bayesian optimization framework on a Riemannian manifold with a global convergence guarantee, to remove the spatio-temporal mismatch between two videos of the same type of activity, and I will also show its outstanding performance on a benchmark in operations research. The last one defines multi-dimensional signal models on both vertices and edges of a dynamic graph, and extracts from imagery social interactions, and I will show its impact on transforming research in social science and education.
About the Speaker: Dr. Ruonan Li is a Research Associate with the School of Engineering and Applied Sciences, Harvard University, where he was a Postdoctoral Fellow. He received the B.Eng. and M.S. degrees with honors in electronic engineering from Tsinghua University, and the Ph.D. degree in electrical engineering from the University of Maryland, College Park. He also held short-term positions at Siemens Corporate Research, Kitware Inc., Comcast Labs, and is visiting to Google's Advanced Technology and Projects group. Ruonan Li's research is focused on understanding signals and datasets that emerge from and evolve with new acquisition and sharing modalities. His work involves image and video data, and is motivated by applications in pattern recognition, semi-supervised and unsupervised learning, spatio-temporal analysis and modeling, and social signal processing.
Robots Perceive, Learn, and Adapt
Speaker:Yu Sun, University of South Florida
Time: 11:00 am - 12:00 pm, Mar 25, 2015
Location: 2MTC, 10th floor, Room 10.099, Brooklyn, NY
To enable robots to work in unstructured environments and perform manipulation tasks with uncertain physical interactions, we propose a novel functional-object-oriented network (FOON) to connect functional-related objects with their interactive manipulation motions. The connections between objects and manipulation motions are learned from observing humans performing daily manipulation tasks. Using a well-trained FOON, robots will be able to decipher a task goal, seek the correct objects to operate, and generate an optimal motion adapted to new conditions. To best facilitate manipulation motions, we also developed new grasping strategies for robots to hold objects with a firm grasp to withstand the disturbance during their physical interactions. In addition, for tasks in computer-assisted surgery, we have developed a comprehensive situation awareness (CSA) platform that is composed of a network of wireless video cameras anchored on the abdominal cavity wall. The CSA platform allows surgeons and computer software to observe a surgical area from different viewpoints and perceive interactions between surgical instrument and tissues for safe and efficient surgery planning and execution.
About the Speaker: Dr. Yu Sun is an Assistant Professor in the Department of Computer Science and Engineering at the University of South Florida. He received his B.S. and M.S. degrees in Electrical Engineering from Dalian University of Technology in 1997 and 2000 respectively, and then Ph.D. degree in Computer Science from the University of Utah in 2007. He serves on several conference/workshop organizing committees, editorial boards, and IEEE RAS governance boards. He recently cofounded the RAS Technical Committee on Robotic Hands, Grasping, and Manipulation and serves as its founding co-chair. His research interests include robotics, intelligence systems, computer vision, virtual reality, human robot interaction, and medical applications.
Energy Storage in Smart Grid: Modeling, Optimization, and Integration
Speaker:Hamed Mohsenian-Rad, University of California, Riverside
Time: 11:00 am - 12:00 pm, Mar 26, 2015
Location: 2MTC, 9th floor, Room 9.009, Brooklyn, NY
The large-scale deployment of batteries and other energy storage technologies is one of the priority areas to build a smart grid, as identified by the U.S. Department of Energy and the National Institute of Standards and Technology. In this presentation, the focus is on addressing the modeling, optimization, and integration challenges in bulk energy storage applications at the power transmission level. It is shown that depending on the MW to GW size of the energy storage systems as well as the grid operating conditions, one can face three different settings for this problem that each carries its unique analytical challenges and requires adequate solution approaches. The practical implications of the presented analytical optimization and integration approaches are discussed based on various case studies using examples from the California power network and energy market. This talk ends by a brief summary of the speaker’s other research projects in the broader area of power systems and smart grid.
About the Speaker: Hamed Mohsenian-Rad received his Ph.D. degree in Electrical and Computer Engineering from the University of British Columbia – Canada in 2008 and his M.Sc. and B.Sc. degrees in Electrical Engineering from Sharif University of Technology – Iran and Amir-Kabir University of Technology – Iran in 2004 and 2002, respectively. He was a post-doctoral fellow at the University of Toronto – Canada from 2009 to 2010. Currently, he is an Assistant Professor of Electrical Engineering at the University of California, Riverside (UCR). At UCR, he is the director of the Smart Grid Research Lab and the Principal Investigator for six major power systems projects, funded by the National Science Foundation, California Energy Commission, and Riverside Public Utilities. He is the recipient of the NSF CAREER Award 2012, the Best paper Award from the IEEE Power and Energy Society General Meeting 2013, and the Best Paper Award from the IEEE International Conference on Smart Grid Communications 2012. Two of his papers are currently the two most cited articles in the IEEE Transactions on Smart Grid. His research interests include modeling, analysis, and optimization of power systems and smart grids with focus on energy storage, renewable power generation, demand response, cyber-physical security, and large-scale power data analysis. Dr. Mohsenian-Rad serves as an Editor for the IEEE Transactions on Smart Grid.
Distributed Coordination of Power Electronics Systems
Speaker:Ali Davoudi, University of Texas-Arlington
Time: 11:00 am - 12:00 pm, Mar 30, 2015
Location: 2MTC, 10th floor, Room 10.009, Brooklyn, NY
Next generation power distribution systems will be massively populated with power electronics devices with ubiquitous computational and communication capabilities to route power among source/load/storage units. Central to realizing this vision are control paradigms that will allow the emergent behaviors of networked power electronics devices to converge to optimal solutions dictated only by mission requirements, and independent of communication structure limitations or individual converter dynamics. Such control frameworks are discussed here for both AC and DC distribution systems. In AC distribution systems, synchronization of parallel inverters and proportional active and reactive load sharing are studied. In DC distribution systems, voltage regulation and proportional load sharing has been achieved. The research outcomes will enable autonomous, self-organizing, and reliable power distribution systems that will be critical building blocks of the emerging smart grid. Current endeavors and future research directions will be discussed.
About the Speaker: Ali Davoudi is currently an Assistant Professor at the Electrical Engineering Department of the University of Texas-Arlington. He received his Ph.D. in Electrical and Computer Engineering from the University of Illinois, Urbana-Champaign, in 2010. His research interests are modeling and control challenges in power electronics systems with applications in renewable energy systems and electrified transportations.
Web-scale Multimedia Data Analytics
Speaker:Zhu Liu, AT&T Labs – Research
Time: 11:00 am - 12:00 pm, Apr 1, 2015
Location: 2MTC, 10th floor, Room 10.009, Brooklyn, NY
Multimedia, the “biggest Big Data,” is becoming the most important and valuable source for information and insights. Ubiquitous multimedia analytics is the indispensable tool that unlocks the underlying value from the explosively increasing amount of multimedia data. While multimedia analytics technologies have found success in many real world applications, enormous challenges, including the scalability, flexibility, efficiency, and reliability of the underlying techniques, still exist. This talk addresses some of these challenges, and it is composed of two subjects. The first topic is on a real-time and robust multimedia content search system, which is composed of audio- and visual-based copy detection modules and a late fusion scheme. The visual module adopts a bag of visual word representation based on the vocabulary tree quantization method, and it is able to query a database of 4.5 million images within 50 milliseconds. The audio module relies on audio fingerprint extracted from sub-band energies, and it takes less than a second to search a database of 6 thousand music tracks. The second topic is about a novel image hashing algorithm powered by deep learning techniques. Traditional image hashing techniques feed hand-crafted visual features into hash functions, and the proposed deep hashing method simultaneously learns both the visual features and the hash functions utilizing a deep network. Comprehensive quantitative evaluations on several large-scale image benchmarks demonstrate its high accuracy in comparison to other supervised hashing algorithms.
About the Speaker: Dr. Zhu Liu is a Principle Inventive Scientist at AT&T Labs – Research. He received the B.S. and M.S. degrees in Electronic Engineering from Tsinghua University, China, in 1994 and 1996, respectively, and the Ph.D. degree in Electrical Engineering from NYU Tandon School of Engineering in 2001. His research interests include multimedia data analytics, computer vision, machine learning, large scale video indexing and retrieval. He authored and co-authored 1 book, 8 book chapters, 12 journal/magazine papers, and more than 50 conference/workshop papers. He also holds 63 issued US patents. Dr. Zhu Liu is on the editorial board of IEEE Signal Processing Letters and the Peer-to-peer Networking and Applications Journal, and he was an associate editor for IEEE Trans. on Multimedia from 2008 to 2012. He has been on organizing committee for many IEEE conferences. Dr. Liu is a senior member of IEEE and a member of ACM.
An Expanding and Expansive View of Computing
Speaker:Jim Kurose, NSF/CISE
Time: 11:00 am - 12:00 pm, Apr 3, 2015
Location: 5 Metrotech Center, Pfizer Auditorium, Brooklyn, NY
Advances in computer and information science and engineering (CISE) are providing unprecedented opportunities for research and education. My talk will begin with an overview of CISE activities and programs at the National Science Foundation and include a discussion of current trends that are shaping the future of our discipline. I will also discuss the opportunities as well as the challenges that lay ahead for our community and for CISE.
About the Speaker: Dr. Jim Kurose is the Assistant Director of the National Science Foundation (NSF) for Computer and Information Science and Engineering (CISE). He leads the CISE Directorate, with an annual budget of more than $850 million, in its mission to uphold the nation's leadership in scientific discovery and engineering innovation through its support of fundamental research in computer and information science and engineering and transformative advances in cyberinfrastructure.
Dr. Kurose is on leave from the University of Massachusetts, Amherst (UMass Amherst), where he has served as Distinguished Professor at the School of Computer Science since 2004. He has also served in a number of administrative roles at UMass and has been a Visiting Scientist at IBM Research, INRIA, Institut EURECOM, the University of Paris, the Laboratory for Information, Network and Communication Sciences, and Technicolor Research Labs.
His research interests include network protocols and architecture, network measurement, sensor networks, multimedia communication, and modeling and performance evaluation. Dr. Kurose has served on many national and international advisory boards and panels and has received numerous awards for his research and teaching. With Keith Ross, he is the co-author of the textbook, Computer Networking, a top down approach (6th edition) published by Addison-Wesley/Pearson.
Dr. Kurose received his Ph.D. in computer science from Columbia University and a Bachelor of Arts degree in physics from Wesleyan University. He is a Fellow of the Association for Computing Machinery (ACM) and the Institute of Electrical and Electronic Engineers (IEEE).
Formulation of Image Quality Metric for Gaussian Blur Images and Dynamic Heart Rate Measurements from Four to Six Seconds Video Frames
Speaker:Raveendran Paramesran, University of Malaya, Malaysia
Time: 2:30 pm - 3:30 pm, Apr 13, 2015
Location: 2MTC, 10th floor, Room 10.009, Brooklyn, NY
In this talk, two current research areas in my lab will be presented. First, we will show how some of the exact Zernike moments can be used in the design of image quality metric to measure the quality of a blurred image. Next, we will show a method on how to obtain the dynamic heart rates of subjects in a cycling exercise from a short duration of video frames.
Features that exhibit human visual perception of image quality scores for blurred images are useful in constructing an image quality metric. In this talk, we show some of the exact Zernike moments (EZMs) that closely model the human quality scores for images of varying degrees of blurriness can be used to measure these distortions. A theoretical framework is developed to identify these EZMs.
The second part of my talk is on how dynamic heart rate measurements that are typically obtained from sensors mounted near to the heart can also be obtained from a short duration of video frames. In this study, two experiments are carried out where a video camera captures the facial images of seven subjects during a cycling exercise.
About the Speaker: P. Raveendran received the B.Sc and M.Sc degrees in electrical engineering from South Dakota State University, Brookings, South Dakota, USA in 1984 and 1985 respectively. He was a systems designer with Daktronics, U.S.A before beginning his academic career as a lecturer at the Department of Electrical Engineering, University of Malaya, Kuala Lumpur, Malaysia in 1986. In 1992, he received a Ronpaku scholarship from Japan to pursue Doctorate in Engineering, which he completed in 1994 at University of Tokushima, Japan. He was promoted to Associate Professor in 1995 and to Professor in 2003. His research areas include image and video analysis, formulation of new image descriptors for image analysis, fast computation of orthogonal moments, analysis of EEG signals, and data modeling of substance concentration acquired from non-invasive methods.
Learning From Distributed Private Data: Algorithms and Applications
Speaker:Anand D. Sarwate, Rutgers University
Time: 11:00 am - 12:00 pm, Apr 17, 2015
Location: LC433, 5 MetroTech Center, Brooklyn, NY
Distributed learning from biomedical data is often hindered by ethical, legal, and technological concerns about data sharing. Data holders wish to maintain control over the uses of their data, and patients or study subjects may be hesitant to allow free and open use of their private medical data. Differential privacy is a framework which allows the quantification of privacy risk. In privacy-preserving distributed learning, the data stays at each site: they locally compute a privacy-preserving summary of their information. The summaries are sent to a private aggregator that performs the final analysis. Differentially private algorithms guarantee privacy by deliberately introducing some noise into the computation – the uncertainty from the noise masks individual data points. This leads to a tradeoff between privacy and accuracy. In this talk I will discuss algorithms for privacy-preserving learning as well as a recent proof-of-concept for this approach applied to neuroimaging data for mental health research.
About the Speaker: Anand D. Sarwate joined as an Assistant Professor in the Department of Electrical and Computer Engineering at Rutgers, the State University of New Jersey in January 2014. He received B.S. degrees in Electrical Engineering and Mathematics from MIT in 2002, an M.S. in Electrical Engineering from UC Berkeley in 2005 and a PhD in Electrical Engineering from UC Berkeleyin 2008. From 2008-2011 he was a postdoctoral researcher at the Information Theory and Applications Center at UC San Diego and from 2011-2013 he was a Research Assistant Professor at the Toyota Technological Institute at Chicago, a philanthropically endowed academic computer science institute located on the University of Chicago campus. He received the NSF CAREER award in 2015.
Control of Islanded Microgrids Encountering Heavy Non-liner Loads
Speaker:Ahmed Mohamed, City University of New York
Time: 2:00 pm - 3:00 pm, Apr 22, 2015
Location: LC433, 5 MetroTech Center, Brooklyn, NY
The smart grid promises radical transformation of the current electricity infrastructure, with a long list of ambitious goals including improved reliability, resiliency and sustainability, smooth integration of renewable and alternative energy resources, controlling emissions, and combating global warming. Microgrid is one of the emerging smart grid killer applications. A truly viable wide-scale microgrid adoption presents utilities with both numerous challenges and opportunities. In this presentation, one of the technical challenges facing islanded microgrids due to their finite inertia will be discussed. Specifically, the impacts of heavy loads, with intermittent non-linear nature, on the stability/security of microgrids operating in an islanded mode will be investigated. A case study on a shipboard power system example representing an islanded microgrid will be presented. The microgrid architecture is assumed to have local renewable energy generation, energy storage system and multiple emergency generators. The impact of non-liner loads will be discussed during normal operating conditions, and in the case of various contingencies. Moreover, a solution to mitigate the effect of these non-linear loads based on storage distribution will be presented.
About the Speaker: Dr. Ahmed Mohamed is an Assistant Professor at the Electrical Engineering Department, City College of the City University of New York (CCNY). He received his Ph.D. degree from Florida International University (FIU), Miami, Florida in 2013. He worked on several federally-funded projects, and have authored/co-authored over 50 peer-reviewed articles/book chapters in a variety of power system fields including smart grids, AC and DC microgrids and renewable energy utilization. He served as a technical reviewer/guest editor on a variety of journals, conferences and IEEE Transactions. He served as a Vice Chair of the IEEE Power and Energy Society (PES), Miami Chapter in 2012 and 2013. He received an “IEEE Outstanding Young Engineer Award” for his services and technical contributions to IEEE Miami Section in 2013.
Wi-Fi in sub-GHz Bands: Research Advances and Global Trials
Speaker:Edward Knightly, Rice University
Time: 11:00 am - 12:00 pm, May 5, 2015
Location: LC400, 5 MetroTech Center, Brooklyn, NY
Worldwide, spectrum regulators are repurposing spectrum and introducing new frameworks for spectrum sharing. In this talk, I will describe new standards, prototypes, and research advances exploiting new spectrum bands. I will focus on the unique characteristics of bands below 1 GHz, often termed the "beach front property" of spectrum due to their superior range and penetration capabilities compared to existing WiFi bands. I will describe capabilities and limitations of recent techniques for realizing high spectral efficiency including medium access exploiting multi-user MIMO. I will draw on experiences from ongoing research trials and measurement studies in Houston, Texas and Itaipu, Brazil.
About the Speaker: Edward Knightly is a professor and the department chair of Electrical and Computer Engineering at Rice University in Houston, Texas. He received his Ph.D. and M.S. from the University of California at Berkeley and his B.S. from Auburn University. He is an IEEE Fellow, a Sloan Fellow, and a recipient of the National Science Foundation CAREER Award. He received best paper awards from ACM MobiCom, IEEE SECON, and the IEEE Workshop on Cognitive Radio Architectures for Broadband. He has chaired ACM MobiHoc, ACM MobiSys, IEEE INFOCOM, and IEEE SECON. He serves as an editor-at-large for IEEE/ACM Transactions on Networking and serves on the IMDEA Networks Scientific Council.
Professor Knightly’s research interests are in the areas of mobile and wireless networks with a focus on protocol design, performance evaluation, and at-scale field trials. He leads the Rice Networks Group. The group’s current projects include deployment, operation, and management of a large-scale urban wireless network in a Houston under-resourced community. This network, Technology For All (TFA) Wireless, is serving over 4,000 users in several square kilometers and employs custom-built programmable and observable access points. The network is the first to provide residential access in frequencies spanning from unused UHF TV bands to legacy WiFi bands (500 MHz to 5 GHz). His group developed the first multi-user beam-forming WLAN system that demonstrates a key performance feature provided by IEEE 802.11ac. His group also co-developed a clean-slate-design hardware platform for high-performance wireless networks, TAPs and WARP.
Improving Long-Term Financial Decision Making Using Behavioral Economics Theory-driven User Interfaces
Speaker:Oded Nov, New York University
Time: 11:00 am - 12:00 pm, May 7, 2015
Location: 2MTC, 10th floor, Room 10.009, Brooklyn, NY
Can human-computer interaction help people make informed and effective decisions about their retirement savings? We applied the behavioral economic theories of endowment effect and loss aversion to the design of novel retirement saving user interfaces. To examine effectiveness, we conducted two experiments in which participants were exposed to one of three experimental user interface designs of a retirement saving simulator, representing endowment effect, loss aversion, social comparison, peer advice and control. Users made 34 yearly asset allocation decisions. We found that designs informed by behavioral economic and social psychology principles and which communicated to savers the long-term implications of their asset allocation choices, led users to adjust their behavior, make larger and more frequent asset allocation changes, and achieve their saving goals more effectively.
Oded Nov is an associate professor at New York University’s Tandon School of Engineering. He received his PhD from Cambridge University. His research focuses on HCI and financial decision making, and on peer production and social computing. Nov is a recipient of the National Science Foundation CAREER Award, and his research is funded by the NSF, the National Academies Keck Initiative, and Google.
New Control Structures for Flexible AC Transmission Systems (FACTS) Devices
Speaker:Saman Babaei, New York Power Authority
Time: 11:00 am - 12:00 pm, May 8, 2015
Location: LC433, 5 MetroTech Center, Brooklyn, NY
The presentation is on the new control structures and converter topologies available to improve the dynamic and steady state performance of the Flexible AC Transmission Systems (FACTS) devices. The combination of the increasing electricity demand and restrictions in expanding the power system infrastructures has urged the deployment of utility-scaled power electronics (FACTS) in the power system. Despite of advances in the semiconductor technology and ultra-fast microprocessor based controllers, there are still many issues to address and room for improvement.
About the Speaker: Saman Babaei is a Research and Technology Development engineer at New York Power Authority (NYPA), R&D department. His work concentrates on power electronics application in power systems, power systems dynamics, renewable energies, and grid energy storage. Pertinent to this presentation, he has worked extensively on the design of new control structures to improve the dynamics performance of the NYPA Convertible Static Compensator (CSC) and some of the results of this work have been published and presented as peer-reviewed journal and conference papers. Saman holds a B.Sc., and a M.Sc., degree all in electrical engineering from Iran University of Science and Technology, Tehran, Iran, and Chalmers University of Technology, Gothenburg, Sweden. He received his PhD in electrical engineering from North Carolina State University, Raleigh, NC, USA in 2014.
Insights from Modeling the Internet’s Social Impact In an Engineering Framework
Speaker:Debasis Mitra, Columbia University
Time: 11:00 am - 12:00 pm, May 19, 2015
Location: LC400, 5 MetroTech Center, Brooklyn, NY
What are the Internet’s essential characteristics, what are the responsible mechanisms and how may these be preserved? To answer some of these questions, we formulate and explore mathematical models, and attempt to draw insights.
We begin with the point of view that Best Effort Service has been an essential contributing factor in the Internet’s explosive growth, and in the spawning of innovations and applications. While there are no QoS guarantees, BE Service has enjoyed reasonable QoS, a low flat subscription fee for broadband connection, and free usage. In the current discussion on Net Neutrality, a central question is whether the ISPs on being allowed to offer Managed Service, will withhold bandwidth for the BE Service with the goal of inducing subscribers to pay a premium price for Managed Service. This would undermine the Internet.
In our models we assume a monopoly ISP which offers both BE Service for free use and MS with guaranteed QoS for a fee per use, and a common portfolio of applications available to customers of both services. Consumers optimize their utility in deciding whether to subscribe to the broadband network, which service (BE or MS) to use, and the usage of the chosen service. The consumers’ decisions depend on the average delay in Best Effort service, which in turn depends on the ISP’s bandwidth allocation to the service. We also assume that BE usage is responsible for the birth of new applications, a point of view commonly asserted in the Net Neutrality debate. We have developed a set of models, which depend on whether the total bandwidth is fixed or extendable through ISP investments, and whether the profit-maximizing ISP is myopic or strategic in its decision-making.
Our results suggest that to preserve a robust offering of BE Service, the regulator may not need to disallow Managed Service. Rather, a strategic ISP will find it in its interest to exploit the power of BE usage-generated applications in conjunction with the MS offering.
Joint work with Qiong Wang, University of Illinois at Urbana-Champaign
About the Speaker: Debasis Mitra joined Columbia University as Professor of Electrical Engineering in 2013. Prior to joining Columbia he worked at Bell Labs for 44 years. His current research interests are in the scientific foundations of policies that impact engineers. Instances are network neutrality, network economics, and the future of industrial laboratories.
Debasis Mitra served as Vice President, Mathematical and Algorithmic Sciences Research Center, in Bell Labs during 1999-2007. He directed work in fundamental mathematics, algorithms, complex systems analysis and optimization, statistics, information & communication sciences and operations research. During 2008-2013 he served as Vice President, Chief Scientist’s Office, Bell Labs.
Debasis Mitra is a member of the National Academy of Engineering, a Bell Labs Fellow and a Life Fellow of the IEEE. He is a recipient of the 2012 ACM SIGMETRICS Lifetime Achievement Award, the 2012 Arne Jensen Lifetime Achievement Award from the International Teletraffic Congress, 1998 IEEE Eric E. Sumner Award, the 1993 Steven O. Rice Prize Paper Award and the 1982 Guillemin-Cauer Prize Paper Award of the IEEE, among other awards.
Debasis Mitra has been on the editorial boards of the IEEE/ACM Transactions on Networking, the IEEE Transactions of Communications, the IEEE Transactions on Circuits and Systems, Queueing Systems (QUESTA) and Operations Research. He is the author of over 100 journal publications and holds over 20 patents. He has served as member, National Academies Panel on Information Sciences (and its predecessors) at the Army Research Laboratory during 2009-2015. In 2011-2012 he chaired the panel and served on the Army Research Laboratory Technical Assessment Board.
Communication in Wireless Networks without a priori CSI: Performance Limits and Communication Strategies
Speaker:Zoran Utkovski, University Goce Delcev in Stip, Macedonia
Time: 11:00 am - 12:00 pm, May 27, 2015
Location: 2MTC, 10th floor, Room 10.099, Brooklyn, NY
Communication services in future wireless networks will probably be delivered by solutions which rely on large antenna arrays and/or network densification by the deployment of small cells. In both cases, the availability of good enough channel state information (CSI) to facilitate phase-coherent processing at multiple antennas or multiple access points appears to be one of the ultimately limiting factors of the performance. This talk presents results for the performance limits of certain wireless networks, which address both the feasibility and the cost associated with the acquisition of channel knowledge. Based on these insights, candidate communication strategies which do not rely on a priori channel state information will be addressed.
About the Speaker: Zoran Utkovski received his Ph.D. (Dr.-Ing) degree (with distinction) from the University of Ulm, Germany in September 2010. He is currently with the University Goce Delcev in Stip, Macedonia, and with the Macedonian Academy of Sciences of Arts. He is recipient of the Fulbright and the DAAD fellowship. His major research interests are in the field of next-generation wireless networks, with the accent on the performance limits and coding schemes for communication in the non-coherent setting (without a priori CSI). He serves as Management Committee Member of the EU COST Actions IC1004 "Cooperative Communications for Green Smart Environments" and IC1101 "Wireless Optical Communications - An Emerging Technology".
Pervasive Personal Robots
Speaker:Ankur Mehta, MIT
Time: 11:00 am - 12:00 pm, Jun 4, 2015
Location: 2MTC, 10th floor, Room 10.099, Brooklyn, NY
Creating and using new robotic systems has typically been limited to experts, requiring engineering background, expensive tools, and considerable time. Instead, I am working to create systems to automatically design, fabricate, and control functional robots from a simple description of the problem at hand. By enabling the on-demand creation of integrated electromechanical systems by casual everyday users, we can get to a point where we can say for any real-world task, "there's a robot for that."
I have moved towards this vision with a system that can easily create programmed printable robots from high-level task descriptions. A software-defined-hardware abstraction allows the algorithmic compilation of fabricable subsystem designs from a structural specification; this is in turn generated from a user assisted grounding of a Structured English behavioral specification. The compiled designs are then manufactured using novel printable manufacturing processes, and programmed with autogenerated code. Advanced wireless protocols and communication hardware enable swarms of such robots to interact with each other and users. In this way, fully functional printable robots can be quickly and cheaply designed, fabricated, and controlled to solve custom tasks by casual users.
About the Speaker: Ankur is a postdoc in the Distributed Robotics group at MIT with Professor Daniela Rus. Pushing towards his vision of a future with robots pervading all aspects of everyday life, his current work focuses on creating a compiler to automatically create custom robots. His system and its generated robots have won recognition as the AFRON Ultra Affordable Educational Robot Project winner and the IROS 2014 best paper.
He completed his graduate degree under Prof. Kris Pister at UC Berkeley as a National Science Foundation Graduate Student Fellow and a Berkeley Fellowship recipient. His PhD work spanned a number of fields, including low power wireless sensor networks, autonomous helicopter and rocket control, and MEMS design.
When not in the lab, Ankur enjoys social dancing, ultimate frisbee, board games, and puzzles.
Modeling of Difference Thresholds of Human Perception
Speaker:Weisi Lin, Nanyang Technological University, Singapore
Time: 11:00 am - 12:00 pm, Jun 11, 2015
Location: 2MTC, 9th floor, Room 9.101, Brooklyn, NY
Just Noticeable Difference (JND) refers to the minimal (critical) amount of “X” that must be changed for the difference to be detectable by humans, where X refers to any signal, derived quantity from signals and even technical specification. JND plays important role explicitly and implicitly throughout our work and life, from sound to smell and from engineering to marketing. "If you cannot measure it, you cannot improve it" (William Thomson); therefore, the scientific measurement and formulation for JND are the prerequisite for user-oriented designs and turning human perceptual (imperfect) sensitivities into system advantages in many circumstances. In this talk, a holistic view will be first presented in JND research and practice, followed by an in-depth case study in visual signals. JND modeling for visual signals has attracted much more research interest so far, while that for audio, haptics, olfaction, gestation and other forms of signals is expected to intensify. In essence, factors to influence JND also include utility, culture and personality, which are to be highlighted as well.
About the Speaker: Weisi Lin received his Ph.D. from King’s College, London University, U.K. He served as the Lab Head of Visual Processing, Institute for Infocomm Research, Singapore. Currently, he is an Associate Professor in the School of Computer Engineering. His areas of expertise include image processing, perceptual signal modeling, video compression, and multimedia communication, in which he has published 120+ journal papers and 200+ conference papers, filed 7 patents, and authored 2 books. He is an AE for IEEE Trans. on Image Processing, IEEE Signal Processing Letters and Journal of Visual Communication and Image Representation, and a past AE for IEEE Trans. on Multimedia. He has also served as a Guest Editor for 8 special issues in international journals. He has been a Technical Program Chair for IEEE ICME 2013, PCM 2012, and QoMEX 2014. He chaired the IEEE MMTC Special Interest Group on QoE (2012-2014). He has been an invited/panelist/keynote/tutorial speaker in 10+ international conferences, as well as a Distinguished Lecturer of Asia-Pacific Signal and Information Processing Association (APSIPA), 2012-2013.
Smartphone Vulnerabilities
Speaker:Alastair Beresford, University of Cambridge
Time: 10:00 am - 11:00 pm, Jun 23, 2015
Location: 2MTC, 9th floor, Room 9.101, Brooklyn, NY
Smartphones today support a large number of applications written by a diverse collection of third-party developers. How secure are the applications and the platforms that support them? What kinds of vulnerabilities exist, and how many phones are vulnerable today? To begin to answer this question, we examine data from the Device Analyzer project, a measurement platform which has collected information from over 23,000 Android phones around the world over the last four years. We find that, on average, 88% of devices were exposed to known privilege-escalation attacks which allow a malicious app to gain root on the device. We also quantify the risk of an alternative attack vector: the JavaScript-to-Java interface vulnerability. This vulnerability allows untrusted JavaScript running in a WebView to break out of the JavaScript sandbox, allowing remote code execution on Android phones. While this vulnerability was first reported in December 2012, we predict that the fix will not be deployed to 95% of devices until January 2018, over 5 years after the release of the fix. The talk finishes with some thoughts on why the security of smartphones is better than the above data might naively suggest, together with some approaches on how we might improve platform security further in the future.
About the Speaker: Alastair Beresford is a Senior Lecturer (aka associate professor) at the University of Cambridge Computer Laboratory. His research work examines the security and privacy of large-scale networked computer systems, and his current focus is on networked mobile devices such as smartphones, tablets and laptops. He looks at the security of the devices themselves as well as the security and privacy problems induced by the interaction between mobile devices and other Internet services.
Empowering Wireless Users with Recommendations and Teleco Providers with Data Analytics
Speaker:Maria Papadopouli, University of Crete
Time: 2:00 pm - 3:00 pm, Jun 24, 2015
Location: 2MTC, 9th floor, Room 9.101, Brooklyn, NY
Wireless access markets become larger, more heterogeneous, and more diverse. Users are differentiated even more by their profile, their demand and quality of experience (QoE) requirements. The large size of these markets and networking environment hinders many analysis and computation complexity challenges. We have developed a multi-layer game-theoretical modeling framework and simulation platform to analyze large-scale markets, networks, and services. The framework applies queuing theory, clustering algorithms, and network economics. Novel aggregation and dimensionality reduction methods address the scalability and accuracy tradeoffs and requirements. Using this framework, we have analyzed the competition in oligopolies, price algorithms, and provisioning of various services (e.g., subscriptions vs. short-term leases).
We have also developed the u-map system, a user-centric reviewing system based on the crowd-sourcing/sensing paradigm. A u-map client enables the smartphone to collect network measurements during a service. Users can also provide opinion scores about their telecommunication services. These objective and subjective measurements are uploaded to a geo-database server. The u-map system models the QoE for different services (e.g., VoIP, video streaming) by applying various advanced machine-learning algorithms. The u-map server provides recommendations to users about wireless operators and service providers. Data analytics can be also applied for churn rate prediction/avoidance, learning about customers, new services planning, and market refinements. The u-map server could run in a cloud-computing environment and can be employed jointly with the modeling framework and simulation platform for better understanding the market, monitoring the infrastructure, and highly efficient resource usage/management. The u-map paradigm has been generalized and used in other domains, such as water distribution networks and medical services/applications. This talk will present the main results of these research activities.
About the Speaker: Maria Papadopouli (Ph.D. Columbia University, October 2002) is an Associate Professor in the Department of Computer Science at the University of Crete and a Research Associate in FORTH-ICS. She has been a visiting Professor at the School of Electrical Engineering, KTH Royal Institute of Technology in Stockholm. Before that, from July 2002 until June 2006, she was a tenure-track Assistant Professor at the University of North Carolina at Chapel Hill (UNC) (on leave from July 2004 until June 2006). Her current research interests are in wireless networking, modeling and performance analysis, network measurements, cognitive radio networks, mobile peer-to-peer computing, and positioning, and pervasive computing.
She has co-authored a monograph on Peer-to-Peer Computing for Mobile Networks: Information Discovery and Dissemination (Springer Eds. 2009). She has served as the co-chair of nine international workshops in the area of wireless networking and mobile peer-to-peer computing and has given more than 35 invited talks in research labs and universities world-wide. In 2004 and 2005, she was awarded with an IBM Faculty Award, while in 2013, she received a Google Faculty Award. In 2012, her research was also funded by the General Secretariat for Research and Technology with an investigator-driven, research excellence grant.
More information about her research can be found at http://www.ics.forth.gr/mobile/
Security-aware Virtual Machine Allocation in the Cloud: A Game Theoretic Approach
Speaker:Charles A. Kamhoua, U.S. Air Force Research Laboratory (AFRL)
Time: 2:00 pm - 3:00 pm, Jun 26, 2015
Location: 2MTC, 9th floor, Room 9.101, Brooklyn, NY
With the growth of cloud computing, many businesses, both small and large, are opting to use cloud services compelled by a great cost savings potential. This is especially true of public cloud computing which allows for quick, dynamic scalability without many overhead or long-term commitments. However, one of the largest dissuasions from using cloud services comes from the inherent and unknown danger of a shared platform such as the hypervisor. An attacker can attack a virtual machine (VM) and then go on to compromise the hypervisor. If successful, then all virtual machines on that hypervisor can become compromised. This is the problem of negative externalities, where the security of one player affects the security of another. This work shows that there are multiple Nash equilibria for the public cloud security game. It also demonstrates that we can allow the players’ Nash equilibrium profile to not be dependent on the probability that the hypervisor is compromised, reducing the factor externality plays in calculating the equilibrium. Finally, by using our allocation method, the negative externality imposed onto other players can be brought to a minimum compared to other common VM allocation methods.
About the Speaker: Charles A. Kamhoua received his B.S. in Electronic from the University of Douala (ENSET), Cameroon in 1999, and the M.S. in Telecommunication and Networking and PhD in Electrical Engineering from Florida International University in 2008 and 2011 respectively. In 2011, he joined the Cyber Assurance Branch of the U.S. Air Force Research Laboratory (AFRL), Rome, New York, as a National Academies Postdoctoral Fellow and became a Research Electronics Engineer in 2012. Prior to joining AFRL, he was an educator for more than 10 years.Dr. Kamhoua is the Principal Investigator of the AFRL in-house basic research project, Survivability through Optimizing Resilient Mechanisms (STORM) funded by the Air Force Office of Scientific Research (AFOSR). He is leading a team of more than 15 researchers including military, summer faculties, postdocs, graduate and undergraduate students from multiple universities across the United States. He is the Program Manager of the first DoD Cyber Security Center of Excellence for HBCU at Norfolk State University in partnership with Tennessee State University and Old Dominium University, a $5 Million project supported by the Air Force, Army, Navy, NSA and the Pentagon. He is also the Program Manager of the Assured Cloud Computing-University Center of Excellence (ACC-UCoE) which is a $6M joint effort of the AFRL Information Directorate, AFOSR, and the University of Illinois at Urbana-Champaign performing state of the art research in cloud security. His managerial and technical expertise is sought from the highest levels within DoD as evidenced by multiple tech transition reviews of DARPA at the Pentagon. His current research interests cover the application of game theory and mechanism design to cyber security and survivability, with over 40 technical publications in prestigious journals and International conferences. He participated in multiple research visits in the United States and abroad to maintain technological excellence in cyber security research relevant to warfighter and civilian needs. His research was presented in multiple national and international conferences and to the Air Force Scientific Advisory Board. He is a reviewer of multiple journals and serves on the technical program committees of several international conferences. He is the Chair of multiple conferences including the 2015 IEEE CLOUD short paper track, the Cloud Computing and Big Data symposium at the 2016 IEEE ICNC, and track chair at the 2015 ICIT. Dr. Kamhoua has been recognized for his scholarship and leadership with numerous prestigious awards including six Air Force Notable Achievement Awards, the 2015 AFOSR Windows on the World Visiting Research Fellowship at Oxford University, UK, an AFOSR basic research award of $645K, the 2015 Black Engineer of the Year Award (BEYA), the 2015 NSBE Golden Torch Award – Pioneer of the Year, a selection to the 2015 Heidelberg Laureate Forum, a National Academies Postdoctoral Fellowship award in 2011, a Best Paper Award at the 2013 IEEE FOSINT-SI, a 2011 NSF PIRE award at Fluminense Federal University, Brazil, and the 2008 FAEDS teacher award. He is an advisor for the National Research Council, a Senior Member of IEEE, a member of ACM, the FIU alumni association, and NSBE.
Wireless 5G: Why, How and What? And Where Are the Opportunities for Us?
Speaker:T. Russell Hsing, National Chiao Tung University, Taiwan
Time: 11:00 am - 12:00 pm, Jul 23, 2015
Location: 2MTC, 9th floor, Room 9.101, Brooklyn, NY
Right after iPhone was released in late June, 2007, Steve Jobs’ Un-equilibrium Relationship occurs immediately (i.e. Create a fascinating market gap of “Demands >> Supply” scenario for consumers Market) and then affect the “Quality of Services” seriously for many wireless operators ultimately. Now the number of wireless subscribers who use smart phones for video streaming applications and high-speed wireless data usage keeps increasing every day. It has been estimated that the traffic volumes will be increased by at least 1000X from 2010 to 2020. The requirement for energy consumptions will also be sharply increased accordingly. Although more and more wireless operators are now starting to deploy the wireless 4G-LTE systems around the globe, but it is clear that the existing technologies just could not fulfill many of newly demanded emerging services which will deal with the new type of social media-enabled traffic pattern and specifically the anticipated data storm ahead of us. It is now the right time for scientists, engineers, technologists, manufacturers and telecom policy makers to get together to have brainstorm on the network architecture, features, functional requirements, business models (e.g.Telecom versus Datacom) and international standards for the future Wireless 5G and wired-wireless convergence networks in order to meet the anticipated demands in our future daily life.
About the Speaker: Based on the speaker’s background and experiences in wireless technologies and networking systems, this talk will describe the reasons WHY we need to have “ wireless 5G” initiated now, HOW we can pursuit it and WHAT will be the final results which we could anticipate to have. A few proposed features (such as Fog-based Radio Access Network, Smart Data Pricing,, Full-Duplex, and SDN-based Vehicular Networks for IoT Applications) for the future Wireless 5G will be addressed. The question of “Where are the opportunities for us in the 5G era” will be answered and T. Russell Hsing is Life Fellow of the IEEE, Fellow of BCS (British Computer Society), and Fellow of SPIE. He is now Chair Professor of National Chiao Tung University in Taiwan, Guest Professor of Peking University in China, Adjunct Professors for Arizona State University (US), Yonsei University (Korea), and Chinese University of Hong Kong. He is also Advisor for the Next Generation Mobile Networks (NGMN) Alliance in Europe, and member of the Advisory Board for the EDGE Lab. of Princeton University and several high-tech start-ups in US and Taiwan. He has been co-Editor-in-Chief (with Prof. Vincent Lau of HKUST and Prof. Mung Chiang of Princeton) of the ICT Book Series for the John Wiley & Sons since 2007. He was Founding Chair of the Vehicular Network and Telematics Applications sub-Technical Committee for the IEEE Communications Society. He has been serving as member of the IEEE Fellow Committee in 2012-2014, the member for the IEEE Eric Sumner Award Committee in 2011-2014. He is now Vice Chair for the IEEE Technical Field Award Council in 2014. His current research areas are: Wireless 5G, Internet of Things (IoT), Software Defined Networks-based Mobility Management, Fog Network & Computing, and Technology Entrepreneurship in academic community.