- talks - lists.cs.princeton.edu

Colloquiym Speaker, Tues April 13th- Wojciech Matusik
by Nicole E. Wagenblast 08 Apr '10

08 Apr '10

Bridging the Gap to the Real Wojciech Matusik, Disney Research Zurich Tuesday, April 13th- 4:30pm Small Auditorium CS105 Devices and systems for measuring real-world shape, motion, and appearance have contributed to progress in computer graphics over the last 15 years. For example, digital photography and image-based rendering have had a profound impact on rendering. Motion capture has transformed computer animation. Geometry acquisition using 3D scanners has driven developments in geometry processing. In data-driven computer graphics, we have replaced hand-modeled and procedurally-defined components with real-world measurements. Furthermore, we have developed tools to modify and edit these data-driven representations. With these new methods in hand, computer graphics has succeeded in recreating reality on a computer screen. The next big challenge for computer graphics and interactive techniques is to expand beyond a computer display. In the next years, the field will be working on algorithms and tools that will allow us to design and automatically fabricate physical output. This trend will dramatically increase the scope of computer graphics and it will exert significant impact on our daily lives. In my talk, I will show two instances of a process that bridges the gap between the digital and real worlds. First, I will describe a mathematical modeling of light reflection from surfaces, based on measurements of physical materials, as well as the reverse process: manufacturing physical materials with desired reflectance properties. Second, I will present an analogous process for measurement, modeling, and fabrication of objects with desired deformation behavior. Not only do these research projects showcase how the scope of computer graphics methods can be greatly extended, but they also prompt fundamental research into the representation and processing of data types that go beyond images and video. Wojciech Matusik is a senior research scientist at Disney Research Zurich. He received a B.S. in EECS from the University of California at Berkeley in 1997, a M.S. in EECS from MIT in 2001, and a Ph.D. in 2003. In 2004, he was named one of the world's top 100 young innovators by MIT's Technology Review Magazine. He received an ACM SIGGRAPH Significant New Researcher Award in 2009. His primary research is in the field of computer graphics, with broad applications in other disciplines such as digital communications, materials science, optics, and biomechanics.

1 0

Colloquium Speaker, Mon April 12- Jake Abernethy
by Nicole E. Wagenblast 06 Apr '10

06 Apr '10

Learning, Adversaries, and Limited Feedback Jake Abernethy, UC Berkeley Monday, April 12, 2010 4:30pm Small Auditorium, CS 105 A basic assumption often made in Machine Learning is that the data we observe are independent and identically distributed. But is this really necessary? Is it even realistic in typical scenarios? A number of recent results provide guarantees for arbitrary (or even adversarially-generated) data sequences. It turns out, for a wide class of problems, the learner's performance is no worse when the data is assumed to be arbitrary as opposed to i.i.d. This result involves a nice application of the Minimax theorem, which I'll briefly describe. I'll also delve into a more challenging problem: the "bandit" setting, in which the learner receives only very little feedback on each example. It was unknown for some time whether there exists an efficient algorithm that achieves the same guarantee for non-i.i.d. data for a general class of learning/decision problems. I'll present the first known such bandit algorithm, and I'll sketch the central ideas behind the technique, borrowing several tricks from the Interior Point optimization literature.

1 0

2pm Fri Apr 9 talk on Content-Centric Opportunistic Communication for Mobile Phones
by Jennifer Rexford 05 Apr '10

05 Apr '10

TALK ANNOUNCEMENT Speaker: Erik Nordstrom, Princeton University Title: Haggle: Content-Centric Opportunistic Communication for Mobile Phones Date/time: 2:00-3:00pm on Friday April 9 Location: room 402 in the Computer Science building Abstract: Today, mobile phones are increasingly used for content-centric communication, whereby users access and share content with each other (e.g., pictures on Facebook). However, content sharing is limited to infrastructure networks and legacy Internet technologies are ill suited for content-centric communication. Moreover, the Internet protocols and APIs do not work well in highly mobile and disruptive environments, where intermittent connectivity is the norm rather than the exception. These deficiencies make it is cumbersome, or even impossible, to share content directly between mobile phones, leading to high stress on existing infrastructure and high costs for the consumer, while putting limitations on the possibilities for developing new and innovative applications. In this talk, I will present Haggle, a novel network architecture built from the ground up to support opportunistic content-centric communication directly between mobile phones. Haggle addresses content rather than hosts, which makes it irrelevant from where a user receives a content item. The content-centric communication is driven by a novel distributed search-based resolution scheme, which draws inspiration from online search engines. The search scheme resolves the mappings between the content and its interest group, i.e., the nodes that have an interest in the content. Search has the benefit of ranking, and therefore provides a natural limitation and ordering mechanism for content dissemination. This means that content, in general, has a higher likelihood of being disseminated if there is a high interest in it in the network. After the talk, I will demonstrate Haggle using a visualization tool that shows how photos taken with a phone's camera are shared with other mobile phones or devices. See http://www.haggleproject.org/ for more details.

1 0

Colloquium Speaker, Tues April 6- Rebecca Fiebrink
by Nicole E. Wagenblast 02 Apr '10

02 Apr '10

Real-time Human-Computer Interaction with Supervised Learning Algorithms for Music Composition and Performance Rebecca Fiebrink, Princeton University, Computer Science Department Tuesday, April 6, 2010- 4:30pm Small Auditorium CS105 Supervised learning offers a useful set of algorithmic tools for many problems in computer music composition and performance. Through the use of training examples, these algorithms offer human musicians a means to implicitly specify the relationship between low-level, human-generated control signals (such as gesturally-manipulated sensor outputs or audio captured by a microphone) and the desired computer response (such as a change in synthesis or structural parameters of dynamically-generated audio). In my work, I explore how to most effectively enable users to interact with supervised learning algorithms to compose and perform new music. I have built a general-purpose software system for applying standard supervised learning algorithms in real-time problem domains. This system, called the Wekinator, supports human interaction throughout the entire supervised learning process, including the generation of training examples and the application of trained models to real-time inputs. Already, the Wekinator has enabled the creation of several new compositions and instruments. Furthermore, this system has enabled me to study several aspects of human-computer interaction with supervised learning in computer music. I have used the Wekinator as a foundation for a participatory design process with practicing composers, ongoing work with non-expert users in a classroom context, and the design of a gesture recognition system for a sensor-augmented cello bow. This research has led to a clearer characterization of the requirements and goals of instrument builders and composers, a better understanding of how to design user interfaces for supervised learning in both real-time and creative application domains, and a greater insight into the roles that interaction (encompassing both human-computer control and computer-human feedback) can play in the development of systems containing supervised learning components. This work highlights how music and other creative endeavors differ from more traditional applications of supervised learning, and it contributes to a broader HCI perspective on machine learning practice.

1 1

C Richards FPO
by Melissa Lawson 01 Apr '10

01 Apr '10

Chris Richards will present his FPO on Thursday April 8 at 11AM in Room 402. The members of his committee are: Andrew Appel, advisor; David Walker and Paul-Andre Mellies (CNRS), readers; David August and Sanjeev Arora, nonreaders. A copy of his thesis, "The Approximation Modality in Models of Higher-Order Types" is on display in Room 310. Everyone is invited to attend his talk.

1 0

Korhammer Lecture: The Art of Green Design: Doing Nothing Well - Prof. Jan M. Rabaey, UC Berkeley
by Jennifer Rexford 25 Mar '10

25 Mar '10

> > > KORHAMMER LECTURE > > PRINCETON UNIVERSITY, DEPARTMENT OF ELECTRICAL ENGINEERING > > “The Art of Green Design: Doing Nothing Well” > > Jan M. Rabaey, Donald O. Pederson Distinguished Professor, University of California, Berkeley > > Monday, March 29, 2010, 4:30pm, Computer Science 104 > > Abstract: > Reducing human appetite for energy and the corresponding carbon footprint is one of the dominant societal problems of today. Information technology (IT) very likely will play a major role in our quest to generate, distribute, store and consume energy more efficiently. Unfortunately the IT platform itself represents a major consumer of electrical power, and its energy footprint is very likely to continue to grow over the next decades. > > The most obvious approach to address this challenge is to continue to improve the energy efficiency of our computation, communication or storage devices. Unfortunately, the opportunities of doing so (using technology scaling or innovative architectures) are getting increasingly limited. Instead, a lower-hanging fruit is to eliminate the massive amount of waste that is inherent to today’s IT systems. Systems tend to be designed and optimized for peak performance. In reality, most computation nodes, networks and storage devices typically operate at a fraction of the maximum load, and do this with surprisingly low energy efficiency. If we could design systems that do “nothing” well (as phrased by David Culler), major energy savings would be enabled. Accomplishing “energy-proportional” computing requires a full-fledged top-down and bottom-up approach to the design of IT systems. This talk discusses what this means for a couple of systems ranging from large scale data centers over communication networks to small-scale body-area networks for medical applications. > > > > >

1 0

Colloquium Speaker- Thurs. March 25- Vinod Vaikuntanathan
by Nicole E. Wagenblast 23 Mar '10

23 Mar '10

Side Channels and Clouds: New Challenges in Cryptography Vinod Vaikuntanathan, IBM T.J. Watson Thursday, March 25, 2010 4:30pm Small Auditorium, CS105 Emerging trends in computation such as cloud computing, virtualization, and trusted computing require that computation be carried out in remote and hostile environments, where attackers have unprecedented access to the devices, the data and the programs. This poses new problems and challenges for cryptography. In this talk, I will present two such challenges, and my recent work towards solving them. 1. Protecting against Side-channel Attacks: Computing devices leak information to the outside world not just through input-output interaction, but through physical characteristics of computation such as power consumption, timing, and electro-magnetic radiation. Such information leakage betrays information about the secrets stored within the devices, and has been successfully utilized to break many cryptographic algorithms in common use. These attacks are commonly called side-channel attacks. Side-channel attacks are particularly easy to carry out when the device is in the physical proximity of an attacker, as is often the case for modern devices such as smart-cards, TPM chips, mobile phones and laptops. In the first part of the talk, I will describe my recent work that lays the foundation of leakage-resilient cryptography - the design of cryptographic schemes that protect against large classes of side-channel attacks. 2. Computing on Encrypted Data: Security in the setting of cloud computing involves a delicate balance of privacy and functionality: while the client must encrypt its data to keep it private from the server, it should also allow for the server to compute on the encrypted data. Can we simultaneously achieve these opposing goals? In the second part of the talk, I will describe an elementary construction of a cryptographic mechanism that allows computation on encrypted data (also called a fully homomorphic encryption scheme). Both these works leverage new mathematical techniques based on geometric objects called lattices. Vinod Vaikuntanathan is a postdoctoral fellow in the cryptography group at IBM T.J. Watson. He received a Ph.D. from MIT in 2009 under the guidance of Shafi Goldwasser. He is a recipient of the MIT Akamai Graduate Fellowship, the IBM Josef Raviv Postdoctoral Fellowship, and more recently, the MIT George M. Sprowls award for the best Ph.D. thesis in Computer Science. The focus of his research is involves the dual goals of devising new mathematical tools for cryptography, as well as applying theoretical cryptography to counter practical attacks.

1 1

Colloquium Speaker, Wed March 24- David Levin
by Nicole E. Wagenblast 22 Mar '10

22 Mar '10

Systems without Cooperation David Levin, University of Maryland Wednesday, March 24 -4:30pm Small Auditorium, CS105 Successful networked systems must account for potentially competing interests: the Internet is no longer the cooperative, technological playground it once was. The protocols are the rules relegating the venue for this competition, but those rules often lack enforcement. I will present the application of economics and trusted hardware to keep participants in a networked system from deviating from the letter and spirit of a protocol. First, I will apply economic mechanism design to keep selfish users from gaining at the expense of others. I will show that the popular BitTorrent system uses, not tit-for-tat as widely believed, but an auction to decide which peers to serve. This model captures known, performance-improving strategies, and shapes or thinking toward new, effective incentive mechanisms. Second, I will apply trusted hardware to keep both selfish and malicious users from "equivocating," or sending semantically conflicting messages. I will present TrInc (Trusted Incrementer), a small piece of trusted hardware intended for use in large-scale distributed systems. With case studies and an implementation, I will demonstrate that TrInc is a practical primitive for protecting a wide range of systems. These two examples together demonstrate the importance of aligning the assumptions of economics and large-scale systems. Doing so allows us to develop new mechanisms that foster cooperation among the otherwise self-interested.

1 1

A Golovinskiy FPO
by Melissa Lawson 15 Mar '10

15 Mar '10

Alex Golovinskiy will present his FPO on Thursday March 18 at 9:30AM in Room 402. The members of his committee are: Tom Funkhouser, advisor; Szymon Rusinkiewicz and Adam Finkelstein, readers; David Dobkin and Rob Schapire, nonreaders. A copy of his thesis, "Analysis of Global Properties of Shapes", is on display in Room 310. Everyone is invited to attend his talk.

1 0

Distinguished Colloquium Speaker Wed. March 10- Barbara Liskov
by Nicole E. Wagenblast 08 Mar '10

08 Mar '10

The Power of Abstraction Barbara Liskov Wednesday, March 10- 4:30pm Friend Convocation Room Abstraction is at the center of much work in Computer Science. It encompasses finding the right interface for a system as well as finding an effective design for a system implementation. Furthermore, abstraction is the basis for program construction, allowing programs to be built in a modular fashion. This talk will discuss how the abstraction mechanisms we use today came to be, how they are supported in programming languages, and some possible areas for future research.

1 1