- talks - lists.cs.princeton.edu

Fwd: EE Departmental Seminar - Enrique Mallada - 3/5/15
by Jennifer Rexford 03 Mar '15

03 Mar '15

> > > Speaker: > Dr. Enrique Mallada > Title: > Distributed Synchronization in Engineering Networks > Day: > Thursday, March 5, 2015 > > > Time: > 4:30 pm > Room: > B205, E-Quad > Host: > Prof. Mung Chiang > > > > > > > > > > Abstract: > This talk presents a systematic study of synchronization on distributed (networked) systems that spans from theoretical modeling and stability analysis to distributed controller design, implementation and verification. We first focus on developing a theoretical foundation for synchronization of networked oscillators. We study how the interaction type (coupling) and network configuration (topology) affect the behavior of a population of heterogeneous coupled oscillators. Unlike existing literature that restricts to specific scenarios, we show that phase consensus (common phase value) can be achieved for arbitrary network topologies under very general conditions on the oscillators' model. > > We then focus on more practical aspects of synchronization on smart grids. We propose a load-side frequency control scheme that can rebalance power and resynchronize frequencies after a disturbance (primary control), while restoring the frequency to its nominal value (secondary control). Unlike the generation-side secondary frequency control that is centralized, our load-side control only requires each bus to communicate with its neighbors. Furthermore, our scheme also provides a fair distribution of the load corrections by minimizing the total disutility of the controllable loads, and is able to preserve inter-area flows and thermal limits. We prove this distributed load-side control is globally asymptotically stable and further illustrate its properties with numerical simulations. > > Bio: > Enrique Mallada is a CMI Post-Doctoral Fellow at California Institute of Technology. He received his Ingeniero en Telecomunicaciones degree from Universidad ORT, Uruguay, in 2005 and his Ph.D. degree in Electrical and Computer Engineering with a minor in Applied Mathematics from Cornell University, Ithaca, in Jan 2014. From 2004 to 2007 he was an IT-Specialist at IBM, and in 2008 he worked as an Engineer at the Traffic Engineering Department of ANTEL. He was also a summer intern in IBM T. J. Watson Research Center during the summer of 2011. Dr. Mallada was awarded the ECE Director's PhD Thesis Research Award for his dissertation in 2014, the Cornell University Jacobs Fellowship in 2011 and the Organization of American States scholarship in the 2008-2009 and 2009-2010 academic years. His research interests include control and dynamical systems, optimization, and networks. > >

1 0

Colloquium Speaker Smita Krishnaswamy Thursday March 5, 12:30pm
by Nicole E. Wagenblast 01 Mar '15

01 Mar '15

Colloquium Speaker Smita Krishnaswamy, from Columbia University Thursday, March 5, 12:30pm Computer Science 105 Unlocking Cellular Computation and Information Processing through Multidimensional Single-Cell Data Cells are computational entities that process external signals through networks of interacting proteins and reconfigure their state via biochemical modifications of proteins and changes in gene expression. Despite progress in the understanding of signaling biology, graph diagrams typically used as depictions of signaling relationships only offer qualitative abstractions. New single-cell measurement technologies provide quantitatively precise measurements of dozens of cellular components representing important biochemical functions. However, a major challenge in deciphering single-cell signaling data is developing computational methods that can handle the complexity, noise and bias in the measurements. I will describe algorithms that quantify the flow of information through signaling interactions and mathematically characterize relationships between signaling molecules, using statistical techniques to detect dependencies while mitigating the effect of noise. I will show how these algorithms can be utilized to characterize signaling relationships in immune cells, detect subtle differences between cell types, and predict differential responses to perturbation. Next, I will analyze T cells from non-obese diabetic (NOD) mice and show that previously recognized defects in extracellular-signal-regulated kinase (ERK) signaling can be traced back to a small receptor-proximal defect that is amplified through reconvergence in the network. Then, I will show how multidimensional extensions of these techniques can be used to track dynamic changes in the relatively unknown network driving the epithelial-to-mesenchymal (EMT) transition that occurs during cancer metastasis, with the goal of predicting drugs to halt the process. Finally, I will discuss future directions involving integration of gene expression and other data types in order to gain a more complete picture of cellular computation.

1 0

Friday March 6: 1:30-2:30 talk on Networks of Networks of Quantum Repeaters
by Margaret Martonosi 27 Feb '15

27 Feb '15

Speaker: Rodney Van Meter, Keio University Title: Networks of Networks of Quantum Repeaters Date/time: Friday March 6, 1:30-2:30pm. Location: CS 401. Abstract: Experimental progress toward quantum repeaters is moving at a tremendous rate, and theorists have proposed half a dozen approaches to managing errors to create high-fidelity entanglement along a chain of repeaters. The next frontier is extending from one-dimensional chains to complex topologies. Problems in network engineering include robust protocol design and resource management. I will give an overview of these issues, then discuss the even more daunting challenge of creating networks of networks -- a true quantum Internet -- capable of coupling networks that are heterogeneous in both physical technology and error management scheme. Bio: Rodney Van Meter received a B.S. in engineering and applied science from the California Institute of Technology in 1986, an M.S. in computer engineering from the University of Southern California in 1991, and a Ph.D. in computer science from Keio University in 2006. His current research centers on quantum computer architecture and quantum networking. Other research interests include storage systems, networking, and post-Moore's Law computer architecture. He is now an Associate Professor of Environment and Information Studies at Keio University's Shonan Fujisawa Campus. Dr. Van Meter is a member of AAAS, ACM and IEEE. Host: Margaret Martonosi (contact me if you’d like to get on his one-on-one schedule.) Margaret Martonosi H. T. Adams '35 Professor of Computer Science. Princeton University ■ mrm(a)princeton.edu ■ http://www.princeton.edu/~mrm

1 0

Colloquium Speaker Sarah Loos Tues March 3, 12:30pm
by Nicole E. Wagenblast 26 Feb '15

26 Feb '15

Colloquium Speaker Sarah Loos, from Carnegie Mellon University Tuesday, March 3, 12:30pm Computer Science 105 Verifying distributed car and aircraft systems with logic and refinement Formal methods and theorem proving have been used successfully in many discrete applications, such as chip design and verification, but computation is not confined to the discrete world. Increasingly, we depend on discrete software to control safety-critical components of continuous physical systems. It is vital that these hybrid (discrete and continuous) systems behave as designed, or they will cause more damage than they intend to fix. In this talk, I will present several challenges associated with verifying hybrid systems and how we use differential dynamic logics to ensure safety for hybrid systems under a continuously infinite range of circumstances and unbounded time horizon. I will discuss the first formal proof of safety for a distributed car control system and the first formal verification of distributed, flyable, collision avoidance protocols for aircraft. In both cases, our verification holds for an unbounded number of cars or aircraft, so the systems are protected against unexpected emergent behavior, which simulation and testing may miss. Providing the first formal proofs of these systems required the development of several seemingly unrelated techniques. We have since identified a unifying challenge for each case study: it is difficult to compare hybrid systems, even when their behaviors are very similar. I will discuss the development of a logic with first-class support for refinement relations on hybrid systems, called differential refinement logic, which aims to address this core challenge for hybrid systems verification within a formal framework. Sarah Loos is a Ph.D. student in the computer science department at Carnegie Mellon University. Her research interests include logical analysis and formal verification of distributed hybrid systems, such as distributed car control and collision avoidance protocols for aircraft. She is a Department of Energy computational science graduate fellow and an NSF graduate research fellow. In addition to her role as co-editor of the ACM-W newsletter, Sarah has served as a student member on the board of trustees and board of advisors for the Anita Borg Institute for Women and Technology. She currently serves on the board of directors for the Foundation for Learning Equality, a non-profit tech company dedicated to bringing the online learning revolution to communities with little or no internet connectivity. Sarah received her undergraduate B.S. degrees in computer science and math from Indiana University.

1 0

Siyu Yang will present his Pre-FPO on March 3, 2015 at 3pm in Room 401.
by Nicki Gotsis 26 Feb '15

26 Feb '15

Siyu Yang will present his Pre-FPO on March 3, 2015 at 3pm in Room 401. His committee members are Professor Robert Tarjan (advisor), Prof Robert Sedgewick (reader), Dr. David Lee (non-Princeton reader), Prof Andrea LaPaugh (non-reader), Prof Bernard Chazelle (non-reader). Everyone is invited to attend his talk. The talk title and abstract follow below: Title: Eager and Lazy Deletion Rebalancing in Binary Search Trees The binary search tree is a fundamental data structure with many applications. Following the invention of AVL tree, many kinds of balanced search trees were developed, with the aim of obtaining simpler updating or improved efficiency or both. In most of such data structures, deletion rebalancing is more complicated than insertion rebalancing. This leads naturally to the question of how much rebalancing is necessary to guarantee a logarithmic height bound. We explore ways to make binary search trees more efficient by delaying rebalancing on deletion. This talk will focus on variants of AVL trees. We propose a variant of weak AVL trees with an even weaker balance condition. Our trees improved the amortized time bounds for rebalancing steps while keeping the same worst-case height guarantee as weak AVL trees. We also propose another variant with lazy rebalancing steps separated from deletion operations themselves. The algorithm's performance is guaranteed by running a background process that rebalances unbalanced parts of the tree. Unlike the similar work of relaxed AVL trees, in which the tree only eventually becomes balanced, our method guarantees that the tree remains balanced at all times, if the background process is run sufficiently often. Finally, I will summarize with some ongoing work.

1 1

Colloquium Speaker Joseph Gonzalez Today, 12:30pm
by Nicole E. Wagenblast 26 Feb '15

26 Feb '15

Colloquium Speaker Joseph Gonzalez, University of California Thursday, February 26, 12:30pm Computer Science 105 Learning Systems: Systems and Abstractions for Large-Scale Machine Learning The challenges of advanced analytics and big data cannot be address by developing new machine learning algorithms or new computing systems in isolation. Some of the recent advances in machine learning have come from new systems that can apply complex models to big data problems. Likewise, some of the recent advances in systems have exploited fundamental properties in machine learning to reach new points in the system design space. By considering the design of scalable learning systems from both perspectives, we can address bigger problems, expose new opportunities in algorithm and system design, and define the new fundamental abstractions that will accelerate research in these complementary fields. In this talk, I will present my research in learning systems spanning the design of efficient inference algorithms, the development of graph processing systems, and the unification of graphs and unstructured data. I will describe how the study of graphical model inference and power-law graph structure shaped the common abstractions in contemporary graph processing systems, and how new insights in system design enabled order-of-magnitude performance gains over general purpose data-processing systems. I will then discuss how lessons learned in the context of specialized graph-processing systems can be lifted to more general data-processing systems enabling users to view data as graph and tables interchangeably while preserving the performance gains of specialized systems. Finally, I will present a new direction for the design of learning systems that looks beyond traditional analytics and model fitting to the entire machine learning life-cycle spanning model training, serving, and management. Joseph Gonzalez is a postdoc in the UC Berkeley AMPLab and cofounder of GraphLab. Joseph received his PhD from the Machine Learning Department at Carnegie Mellon University where he worked with Carlos Guestrin on parallel algorithms and abstractions for scalable probabilistic machine learning. Joseph is a recipient of the AT&T Labs Graduate Fellowship and the NSF Graduate Research Fellowship.

1 0

Colloquium Speaker Joseph Gonzalez Thurs Feb 26, 12:30pm
by Nicole E. Wagenblast 25 Feb '15

25 Feb '15

Colloquium Speaker Joseph Gonzalez, University of California Thursday, February 26, 12:30pm Computer Science 105 Learning Systems: Systems and Abstractions for Large-Scale Machine Learning The challenges of advanced analytics and big data cannot be address by developing new machine learning algorithms or new computing systems in isolation. Some of the recent advances in machine learning have come from new systems that can apply complex models to big data problems. Likewise, some of the recent advances in systems have exploited fundamental properties in machine learning to reach new points in the system design space. By considering the design of scalable learning systems from both perspectives, we can address bigger problems, expose new opportunities in algorithm and system design, and define the new fundamental abstractions that will accelerate research in these complementary fields. In this talk, I will present my research in learning systems spanning the design of efficient inference algorithms, the development of graph processing systems, and the unification of graphs and unstructured data. I will describe how the study of graphical model inference and power-law graph structure shaped the common abstractions in contemporary graph processing systems, and how new insights in system design enabled order-of-magnitude performance gains over general purpose data-processing systems. I will then discuss how lessons learned in the context of specialized graph-processing systems can be lifted to more general data-processing systems enabling users to view data as graph and tables interchangeably while preserving the performance gains of specialized systems. Finally, I will present a new direction for the design of learning systems that looks beyond traditional analytics and model fitting to the entire machine learning life-cycle spanning model training, serving, and management. Joseph Gonzalez is a postdoc in the UC Berkeley AMPLab and cofounder of GraphLab. Joseph received his PhD from the Machine Learning Department at Carnegie Mellon University where he worked with Carlos Guestrin on parallel algorithms and abstractions for scalable probabilistic machine learning. Joseph is a recipient of the AT&T Labs Graduate Fellowship and the NSF Graduate Research Fellowship.

1 0

Colloquium Speaker Suman Jana today, 12:30pm
by Nicole E. Wagenblast 25 Feb '15

25 Feb '15

Colloquium Speaker Suman Jana, Stanford University Wednesday, February 25, 2015 - 12:30pm Computer Science 105 Rise of the Planet of the Apps: Security and Privacy in the Age of Bad Code Computing is undergoing a major shift. Third-party applications hosted in online software markets have become ubiquitous on all kinds of platforms: mobile phones, Web browsers, gaming devices, even household robots. These applications often include yet more third-party code for advertising, analytics, etc. These trends have dramatically increased the amount of bad code throughout the software stack - buggy code, malicious code, code that overcollects private information intentionally or by accident, overprivileged code vulnerable to abuse - as well as the amount of sensitive data processed by bad code. In this talk, I will demonstrate that existing application platforms are ill-suited to dealing with bad code, thus causing security and privacy problems. I will then show how to isolate bad code without affecting its useful functionality, by redesigning the interfaces across the software stack and controlling the information released to the applications by the platform. I will also show how automated testing can identify bad code and help developers improve their applications. Suman Jana is a postdoctoral researcher at Stanford University. He earned his PhD in 2014 from the University of Texas, where he was supported by the Google PhD Fellowship. He is broadly interested in identifying fundamental flaws in existing systems and building new systems with strong security and privacy guarantees. Suman received the 2014 PET Award for Outstanding Research in Privacy-Enhancing Technologies, Best Practical Paper Award from the 2014 IEEE Symposium on Security and Privacy (Oakland), Best Student Paper Award from the 2012 IEEE Symposium on Security and Privacy, and the 2012 Best Applied Security Paper Award. Suman's research has been widely covered in popular media, and his code has been deployed at Google, Mozilla, and Apache.

1 0

Correction- Colloquium Speaker Joseph Gonzalez Thurs Feb 26, 12:30pm
by Nicole E. Wagenblast 24 Feb '15

24 Feb '15

Colloquium Speaker Joseph Gonzalez, University of California Thursday, February 26, 12:30pm Computer Science 105 Learning Systems: Systems and Abstractions for Large-Scale Machine Learning The challenges of advanced analytics and big data cannot be address by developing new machine learning algorithms or new computing systems in isolation. Some of the recent advances in machine learning have come from new systems that can apply complex models to big data problems. Likewise, some of the recent advances in systems have exploited fundamental properties in machine learning to reach new points in the system design space. By considering the design of scalable learning systems from both perspectives, we can address bigger problems, expose new opportunities in algorithm and system design, and define the new fundamental abstractions that will accelerate research in these complementary fields. In this talk, I will present my research in learning systems spanning the design of efficient inference algorithms, the development of graph processing systems, and the unification of graphs and unstructured data. I will describe how the study of graphical model inference and power-law graph structure shaped the common abstractions in contemporary graph processing systems, and how new insights in system design enabled order-of-magnitude performance gains over general purpose data-processing systems. I will then discuss how lessons learned in the context of specialized graph-processing systems can be lifted to more general data-processing systems enabling users to view data as graph and tables interchangeably while preserving the performance gains of specialized systems. Finally, I will present a new direction for the design of learning systems that looks beyond traditional analytics and model fitting to the entire machine learning life-cycle spanning model training, serving, and management. Joseph Gonzalez is a postdoc in the UC Berkeley AMPLab and cofounder of GraphLab. Joseph received his PhD from the Machine Learning Department at Carnegie Mellon University where he worked with Carlos Guestrin on parallel algorithms and abstractions for scalable probabilistic machine learning. Joseph is a recipient of the AT&T Labs Graduate Fellowship and the NSF Graduate Research Fellowship.

1 0

Colloquium Speaker Joseph Gonzalez Thurs Feb 24, 12:30pm
by Nicole E. Wagenblast 24 Feb '15

24 Feb '15

Colloquium Speaker Joseph Gonzalez, University of California Thursday, February 24, 12:30pm Computer Science 105 Learning Systems: Systems and Abstractions for Large-Scale Machine Learning The challenges of advanced analytics and big data cannot be address by developing new machine learning algorithms or new computing systems in isolation. Some of the recent advances in machine learning have come from new systems that can apply complex models to big data problems. Likewise, some of the recent advances in systems have exploited fundamental properties in machine learning to reach new points in the system design space. By considering the design of scalable learning systems from both perspectives, we can address bigger problems, expose new opportunities in algorithm and system design, and define the new fundamental abstractions that will accelerate research in these complementary fields. In this talk, I will present my research in learning systems spanning the design of efficient inference algorithms, the development of graph processing systems, and the unification of graphs and unstructured data. I will describe how the study of graphical model inference and power-law graph structure shaped the common abstractions in contemporary graph processing systems, and how new insights in system design enabled order-of-magnitude performance gains over general purpose data-processing systems. I will then discuss how lessons learned in the context of specialized graph-processing systems can be lifted to more general data-processing systems enabling users to view data as graph and tables interchangeably while preserving the performance gains of specialized systems. Finally, I will present a new direction for the design of learning systems that looks beyond traditional analytics and model fitting to the entire machine learning life-cycle spanning model training, serving, and management. Joseph Gonzalez is a postdoc in the UC Berkeley AMPLab and cofounder of GraphLab. Joseph received his PhD from the Machine Learning Department at Carnegie Mellon University where he worked with Carlos Guestrin on parallel algorithms and abstractions for scalable probabilistic machine learning. Joseph is a recipient of the AT&T Labs Graduate Fellowship and the NSF Graduate Research Fellowship.

1 0