- talks - lists.cs.princeton.edu

B Guo preFPO
by Melissa M Lawson 02 Jan '07

02 Jan '07

Bolei Guo will present her preFPO on Tuesday January 9 at 1PM in Room 402. The members of her committee are; David August, advisor; Andrew Appel and Michael Hind (IBM TJ Watson), readers; Brian Kernighan and David Walker, nonreaders. Everyone is invited to attend her talk. Her abstract follows below. ------------------------------------------- "Low-Level Shape Analysis with Inductive Recursion Synthesis" Aggressive memory disambiguation for low-level code can eliminate the problem of conservatively propagating high-level memory analysis results. It critically affects the effective-ness of back-end code optimizations. In particular, automatic loop parallelization relies on a type of memory analysis called shape analysis, which determines the shapes of data structures built by the program. We designed a separation-logic-based shape analysis for low-level code. Separation logic with recursively defined predicates allows for concise yet precise description of the shapes of data structures. However, inferring such predicates is a difficult task. Current uses of separation logic involve pre-defined built-in predicates and logic rules specificially written for these predicates, limiting the class of analyzable programs. We developed an inductive learning technique to automatically infer arbitrary recursive shape predicates. In addition, to facilitate the interaction between local reasoning and the global invariants given by the recursive predicates, a notion of "truncation points" is introduced in a predicate, which gives rise to generic algorithms for unfolding and folding arbitrary data structures. This shape analysis uses the alias information provided by a fast low-level pointer analysis to identify recursive data structures and to prune away code that does not affect the result of shape analysis. It has been implemented in the Velocity compiler and evaluated on benchmarks from the SPEC2000 and Olden benchmark suites.

1 0

J Podolak preFPO
by Melissa M Lawson 11 Dec '06

11 Dec '06

Joshua Podolak will present his preFPO on Thursday December 14 at 3:30PM in Room 402. The members of his committee are Szymon Rusinkiewicz, advisor; Adam Finkelstein and Tom Funkhouser, readers; David Dobkin and Bernard Chazelle, nonreaders. Everyone is invited to attend his talk. His abstract follows below. ------------------------------ Symmetry is an important feature of almost all shapes: nearly every man-made object contains at least one plane of perfect symmetry and most natural objects exhibit near-perfect symmetries. However, symmetry information has been underutilized in computer graphics. In this thesis we will analyze both the perfect and the imperfect symmetries of 2D and 3D models and define a symmetry transform that captures all planar symmetries. We then examine some applications that can benefit from symmetry information such as alignment, completion, shape matching, segmentation, automatic viewpoint selection and remeshing.

1 0

Of possible interest in COS 576: Ron Weiss talk
by Kenneth Steiglitz 05 Dec '06

05 Dec '06

This Thursday Ron Weiss will be a guest speaker in COS 576 (Room 105) -------------------------------------------------- Ron Weiss, Dept. EE Synthetic Biology and Genetic Computing Thursday Dec. 7, 3pm, CS Room 105 (small auditorium) -------------------------------------------------- All welcome, ken

1 0

PICASso TODAY: Analyzing Protein Interaction Networks
by Chrissy Colvin 29 Nov '06

29 Nov '06

** PICASso: ** Program in Integrative Information, Computer and Application Sciences ** www.cs.princeton.edu/picasso ** ** TALK TODAY--Wednesday, November 29, 2006 Computation and Data Analysis in Biology and Information Sciences www.cs.princeton.edu/picasso/bio_info_lunch.php TITLE: Analyzing Protein Interaction Networks SPEAKER: Mona Singh, Computer Science, Lewis-Sigler Institute, Princeton University TIME: TODAY Wednesday, November 29, 2006 Seminar begins at 12:30 p.m. (lunch provided 12:20) LOCATION: Room 402 Computer Science, Princeton University ABSTRACT: High-throughput technologies have determined physical interactions for a significant fraction of proteins in several organisms. Taken together, these linkages form large protein interaction networks. Computational analysis of these networks can help uncover the functions of individual proteins as well their roles within larger pathways. In this talk, I will introduce novel methods for analyzing protein interaction networks and uncovering recurring means with which diverse biological processes are carried out. Our approach lends insight into the organizational principles of protein networks and can be used to determine functionally cohesive subnetworks and help predict protein function. SIGN UP FOR THE PICASSO MAILING LIST: ===================================== If you would like to be kept informed of computationally-oriented events in (and around) Princeton, please SUBSCRIBE to the PICASso mailing list by visiting https://lists.cs.princeton.edu/mailman/listinfo/picasso. This page also contains information on how to UNSUBSCRIBE. PLEASE FORWARD THIS MESSAGE TO OTHER COMPUTATIONALLY-ORIENTED RESEARCHERS WHO MAY BE INTERESTED IN THIS EVENT, OR FUTURE PROGRAMS. THANKS!

1 0

D Nehab preFPO
by Melissa M Lawson 27 Nov '06

27 Nov '06

Diego Nehab will present his preFPO on Thursday November 30 at 4PM in Room 401 (note room!). The members of his committee are: Szymon Rusinkiewicz, advisor, James Davis (UC Santa Cruz), Tom Funkhouser, Adam Finkelstein, and Kai Li. Everyone is invited to attend his talk. His abstract appears below. --------------------------- In this talk, we present a variety of results that push the state of the art in 3D scanning techniques towards new simplicity, generality, and precision levels. We start with the introduction of the temporal stereo triangulation scanner, which is composed of two cameras and a projector. The system is substantially simpler than previous designs, because the projector does not need to be calibrated or even synchronized with the stereo camera pair. In order to improve the quality of our scanner, we introduce a novel algorithm that performs symmetric sub pixel refinement of integer matches. By simultaneously updating the matching coordinates on each camera, the method avoids biases that would otherwise appear as coherent noise in the final reconstructions. We then investigate the combination of data coming from a variety of different scanner types. We notice that the noise properties of acquired normal fields are generally unrelated to those commonly present in triangulated data. By carefully combining the best portion of each measurement type, we are able to achieve substantially higher quality results. Both normal acquisition and stereo triangulation are usually designed under the assumption of diffuse, opaque materials. We therefore conclude the talk with an investigation of alternative techniques for acquiring shape and orientation from translucent and/or highly specular materials.

1 0

Y Wang preFPO
by Melissa M Lawson 27 Nov '06

27 Nov '06

Yong Wang will present his preFPO on Thursday November 30 at 10:30AM in Room 402. The members of his committee are: Margaret Martonosi, advisor; Li-Shiuan Peh, Jennifer Rexford, Larry Peterson, and Kai Li. Everyone is invited to attend his talk.. His abstract follows below. ------------------------------ Wireless sensor networks and disruption-tolerant networks are new types of challenged networks that have very unpredictable and unusual situations, quite distinct from what a protocol architecture is usually designed for. The unique communication characteristics of these networks have created new, interesting challenges for the research community, such as opportunistic mobility, sparse connectivity, frequent network disruption/disconnection, and severely constrained energy budgets. Such challenges demand novel solutions to maintain efficient routing and take advantage of existing network services. This requires the network to be aware of such challenges and respond appropriately. In this thesis, I explore new optimizations via situation-awareness to improve routing performance in the presence of such challenges and disruptions. In particular, I study five categories of challenges that cover a wide range of disruptions experienced in challenged networks: sparse end-to-end connectivity, highly varying mobility, opportunistic connectivity, intermittent connectivity, and network congestion. The first three categories are confined to mobile networks, while the last two categories apply to static networks. The first category consists of networks with sparse end-to-end connectivity. Therefore, route breakage may bring a communication pair into temporary disconnection or lead to nodes using stale routes. I propose a timekeeping technique to drive route prefetching and decay operations based on route lifetime information. Through extensive simulations, I demonstrate up to 31% improvement in packet delivery rate and 53% improvement in average end-to-end packet latency. The second category consists of networks with a varying mobility pattern that is typical in many emerging networks, an example of which is ZebraNet. With such a mobility pattern, nodes move in phases that have very different characteristics I propose a model-based approach to capture mobility phase changes in order to maintain efficient routing. When evaluated using a mobility trace from the real-world ZebraNet deployment, the situation-aware approach leads to an improvement of up to 120% in packet delivery rate. _______________________________________________ talks mailing list talks(a)lists.cs.princeton.edu https://lists.cs.princeton.edu/mailman/listinfo/talks

1 0

Y Wang preFPO
by Melissa M Lawson 27 Nov '06

27 Nov '06

Young Wang will present his preFPO on Thursday November 30 at 10:30AM in Room 402. The members of his committee are: Margaret Martonosi, advisor; Li-Shiuan Peh, Jennifer Rexford, Larry Peterson, and Kai Li. Everyone is invited to attend his talk.. His abstract follows below. ------------------------------ Wireless sensor networks and disruption-tolerant networks are new types of challenged networks that have very unpredictable and unusual situations, quite distinct from what a protocol architecture is usually designed for. The unique communication characteristics of these networks have created new, interesting challenges for the research community, such as opportunistic mobility, sparse connectivity, frequent network disruption/disconnection, and severely constrained energy budgets. Such challenges demand novel solutions to maintain efficient routing and take advantage of existing network services. This requires the network to be aware of such challenges and respond appropriately. In this thesis, I explore new optimizations via situation-awareness to improve routing performance in the presence of such challenges and disruptions. In particular, I study five categories of challenges that cover a wide range of disruptions experienced in challenged networks: sparse end-to-end connectivity, highly varying mobility, opportunistic connectivity, intermittent connectivity, and network congestion. The first three categories are confined to mobile networks, while the last two categories apply to static networks. The first category consists of networks with sparse end-to-end connectivity. Therefore, route breakage may bring a communication pair into temporary disconnection or lead to nodes using stale routes. I propose a timekeeping technique to drive route prefetching and decay operations based on route lifetime information. Through extensive simulations, I demonstrate up to 31% improvement in packet delivery rate and 53% improvement in average end-to-end packet latency. The second category consists of networks with a varying mobility pattern that is typical in many emerging networks, an example of which is ZebraNet. With such a mobility pattern, nodes move in phases that have very different characteristics I propose a model-based approach to capture mobility phase changes in order to maintain efficient routing. When evaluated using a mobility trace from the real-world ZebraNet deployment, the situation-aware approach leads to an improvement of up to 120% in packet delivery rate.

1 0

PLOrk performance tonight
by Ge Wang 21 Nov '06

21 Nov '06

Dear all, This event won't be a talk (we hope), but it seemed more relevant to announce here than to [building] or [downtime]. http://ffmup.org/nov_21_06.html --- who/what: The Princeton Laptop Orchestra (PLOrk) presents an evening of new works for computer-mediated chamber music (duo/trio/quartet), composed and performed by members of the Fall 2006 ensemble and seminar: Alan Tormey | Ananya Misra | Andrea Mazzariello | Anne Hege | Ge Wang | John Fontein | Laurie Hollander | Matt Hoffman | Margaret R. Daniel | Michael Hammond | Perry Cook | Rebecca Fiebrink | Samson Young | Scott Elmegreen | Scott Smallwood | Spencer Salazar --- when: tonight 9:00pm (Tuesday) where: Terrace Club (free admission) why: there is usually free beer at ffmup events; new music; free beer! Hope to see you there! Best, Ge!

1 0

Michael J Freedman, Mon Nov 20, 1:30
by Vivek Pai 16 Nov '06

16 Nov '06

Michael J Freedman, the NYU graduate student, will be speaking on Monday at 1:30 in room 402. His talk title and abstract are included below. If you'd like to talk with him, please sign up in /n/fs/department/schedules/visitors/visitfall06/freedman Just to avoid some confusion that occurred the last time he visited, please note that this is not Michael H. Freedman, the Fields Medal winner. =================== RE: Fighting Spam with Reliability and Privacy Mike Freedman NYU / Stanford http://www.scs.stanford.edu/mfreed/ mfreed(a)scs.stanford.edu ----- The explosive growth in unwanted email has prompted the development of techniques for the rejection of email, intended to shield recipients from the onerous task of identifying the legitimate email in their inboxes amid a sea of spam. Unfortunately, widely-used content-based filtering systems have converted the spam problem into a false positive one: Email has become unreliable. We present Reliable Email (RE: [1]), a new whitelisting system that incurs zero false positives among socially-connected users yet populates whitelists automatically by exploiting transitive trust relationships. Further, while RE: permits a recipient to discover whether other users have whitelisted the sender, it preserves the privacy of email contacts with a cryptographic Private Matching [2] protocol. Using real email traces from two sites, we demonstrate that RE: renders a significant fraction of received email reliable. Our evaluation also shows that RE: can prevent up to 88% of the false positives incurred by common email rejection systems. Finally, we propose a more formal definition of security for verifying social connectedness in a privacy-preserving fashion [3]. In the process, we identify a mismatch between RE:'s goals and [2]'s properties, enabling us to construct new cryptographic protocols that are either more efficient (based only on symmetric operations) or achieve stronger guarantees (namely, non-transferability) through bilinear pairings. Using such, RE: can provide privacy-preserving social whitelisting with millisecond overhead. 1. Re: Reliable Email Garriss, Kaminsky, Freedman, Karp, Mazieres, and Yu, NSDI '06 2. Efficient Private Matching and Set Intersection Freedman, Nissim, and Pinkas, EUROCRYPT '04 3. Efficient Private Techniques for Verifying Social Proximity Freedman and Nicolosi, Submitted Nov '06 ----- Michael J. Freedman is a doctoral student at NYU, currently visiting Stanford University. He received his M.Eng. and S.B. degrees from MIT. His research interests broadly focus on security, distributed systems, and cryptography. Among other things, he developed the Coral content distribution network and OASIS anycast service, which serve more than a million users daily (see www.coralcdn.org). In 2006, he co-founded Illuminics Systems, which has been acquired by Quova, Inc.

1 0

Gödel Centenary Program this Friday
by Moses Charikar 15 Nov '06

15 Nov '06

On Friday, November 17, from 2:00 p.m. to 6:30 p.m. in Wolfensohn Hall, the Institute will host A Program to Mark the Centenary Year of the Birth of Kurt Gödel. The event is free and open to the public; reservations are not required. Seating is on a first come, first served basis. The program will consist of talks aimed at a general audience on the life and work of Kurt Gödel (1906-1978) and his impact on mathematics, philosophy, and computer science. Participants include: John Burgess, Princeton University; John W. Dawson, Jr., The Pennsylvania State University; Solomon Feferman, Stanford University; Juliette Kennedy, University of Helsinki; Karl Sigmund, University of Vienna; and Avi Wigderson, Institute for Advanced Study. For additional information about the program, please visit http://www.ias.edu/SpFeatures/kurt_godel/godel-4.php. There is also background on Gödel's life and work accessible from the IAS homepage, www.ias.edu. An exhibition of posters on Gödel, created by the University of Vienna, will remain on view on the main floor of Fuld Hall through Sunday, December 3. --Moses

1 0