- talks - lists.cs.princeton.edu

Kavita Bala talk
by Tom Funkhouser 17 Sep '08

17 Sep '08

*Speaker: Kavita Bala, Cornell University Title: Perceptually-based scalable graphics Abstract: Simulating the appearance of complex scenes faithfully and efficiently is a challenge in graphics. My research develops algorithms that scale to complex illumination and scenes, by exploiting limitations in human perception. In this talk I will describe two complementary research goals. First, we need to understand image fidelity: when is a rendered image good enough? We introduce a new appearance-based measure of image fidelity called visual equivalence[SIG07,SIG08] that goes beyond pixel accuracy to better capture what graphics practitioners care about: preserving appearance in complex scenes. Second, we design perceptually-based algorithms that scale to complex scenes and complex illumination effects, like motion blur, depth-of-field, participating media, and subsurface scattering. I will describe our work on lightcuts for scalable final rendering [SIG05,SIG06,EG08] and matrix row-column sampling for lighting preview [SIG07,EGSR08]. These approaches are complementary in performance and can be applied across a range of applications such as cinematic relighting, production rendering, games, cultural heritage, and ecommerce, among others. Bio: Kavita Bala is an Assistant Professor in the Computer Science Department and Program of Computer Graphics at Cornell University. She received her S.M. and Ph.D. from the Massachusetts Institute of Technology (MIT), and her B.Tech. from the Indian Institute of Technology (IIT, Bombay). Bala specializes in computer graphics, leading research projects in scalable rendering, perceptually-based rendering, interactive global illumination, and image-based modeling and texturing. She has co-authored the graduate-level textbook "Advanced Global Illumination" (A K Peters publisher, second edition). In 2005 she co-chaired the Eurographics Symposium on Rendering (EGSR). Bala has received the NSF CAREER award, Cornell's College of Engineering James and Mary Tien Excellence in Teaching Award, and Cornell's Affinito-Stewart award. **Time/Place: Monday, September 22, 2008 12:00 - 1:30PM, CS 402 (PIXL Lunch) *

1 0

E Banks preFPO
by Melissa Lawson 16 Sep '08

16 Sep '08

Eric Banks will present his preFPO on Monday September 22 at 1:30PM in room 402. The members of his committee are: Mona Singh, advisor; Olga Troyanskaya and Bernard Chazelle, readers; Tom Funkhouser and David Blei, nonreaders. Everyone is invited to attend his talk. His abstract follows below. ------------------------------------------ High-throughput experimental and computational approaches are enabling the collection of large-scale biological networks, consisting of proteins and the interactions among them, for a growing number of species. With appropriate computational analysis and experimental work the potential exists for uncovering the organizational principles of the cell and, consequently, protein functions and pathways, which are still largely unidentified. In this work, we introduce a novel framework for analyzing protein interaction networks in order to uncover organizational units corresponding to recurring means with which diverse biological processes are carried out. We formalize recurring patterns of interaction among different types of proteins using "network schemas"; network schemas specify descriptions of proteins and the topology of interactions among them. We develop algorithms for systematically uncovering recurring, over-represented schemas in physical interaction networks and apply these methods to the S. cerevisiae interactome, identifying hundreds of such organizational units of varying complexity. We establish the functional importance of these schemas by showing that they correspond to functionally cohesive sets of proteins, are enriched in the frequency with which they have instances in the H. sapiens interactome, and are useful for predicting protein function. Finally, we introduce NetGrep, a system for searching protein interaction networks for matches to more general network schemas. NetGrep provides an advanced graphical interface for specifying schemas and fast algorithms for extracting their matches.

1 0

Matthew Bridges FPO
by Melissa Lawson 09 Sep '08

09 Sep '08

Matt Bridges will present his FPO on Monday September 15 at 2PM in Room 402. The members of his committee are: David August, advisor; Kai Li and Dan Lavery, readers; Doug Clark and Margaret Martonosi, nonreaders. A copy of his thesis, "The VELOCITY Compiler: Extracting Efficient Multicore Execution from Legacy Sequential Codes", is on display in Room 310. Everyone is invited to attend his talk.

1 0

F Perry FPO
by Melissa Lawson 21 Aug '08

21 Aug '08

Frances Perry will present her FPO on Thursday August 28 at 10:30AM in room 402. The members of her committee are: David Walker, advisor; Andrew Appel and David Tarditi, readers; David August and Margaret Martonosi, nonreaders. A copy of her thesis, "Reasoning About Software in the Presence of Transient Faults", is on display in room 310. Everyone is invited to attend her talk.

1 0

C DeCoro preFPO
by Melissa Lawson 20 Aug '08

20 Aug '08

Chris DeCoro will present his preFPO on Wednesday August 27 at 2PM in Room 402. The members of his committee are Szymon Rusinkiewicz (advisor); Tom Funkhouser and Tim Weyrich (readers); and Fei-Fei Li and Brian Kernighan (non-readers). A copy of his abstract follows below. Everyone is invited to attend his talk. =========================== Title: "Filters for Detail Control and Stylization in Rendering" This dissertation examines a new class of filtering operations that are applied during rendering -- the process of synthesizing an image from geometry. These ``rendering filters'' allow for detail control, creative stylization, and noise rejection in ways not possible with purely image- or geometry-space filters. In particular: the subtractive shadow filter enables flexible control of shadowing level-of-detail in a real-time animation, allowing designers to rapidly prototype changes to materials and geometry. In our system, direct illumination is always computed and preserved in full detail. In contrast, the degree to which the perceptual cues provided by shadowing are maintained is controlled by the user, as a tradeoff between accuracy and rendering speed. Additionally, the subtractive shadow filter is able to take advantage of the low-frequency content of shadows, relative to direct illumination, to improve perceptual quality relative to conventional shadow computation with a limited number of lights. The stylized illumination filter gives the digital artist creative control over the appearance of shadows and illumination, replicating many of the effects frequently employed by traditional artists; In particular, we give the user control over the key parameters of inflation, brightness, softness, and abstraction. Because the filter operates in real time, the artist may design the desired effect with interactive feedback. The path-density filter allows for rejection of statistical outliers in the computation of global illumination solutions, significantly decreasing noise and improving convergence. The key insight of the method is to leverage density-estimation techniques in a high-dimensional space representing scattering events along light paths; paths in low-density regions are considered outliers. Unlike existing methods, it does so in a way that is both robust to large-scale noise, while efficient on well-behaved distributions.

1 0

Z Barutcuoglu FPO
by Melissa Lawson 19 Aug '08

19 Aug '08

Zafer Barutcuoglu will present his FPO on Monday August 25 at 1PM in Room 402. The members of his committee are: Rob Schapire, advisor; Olga Troyanskaya and David Blei, readers; Mona Singh and Fei-Fei Li, nonreaders. A copy of his thesis, "Using Structural Information in Machine Learning Applications", is on display in Room 310. Everyone is invited to attend his talk.

1 0

G Ottoni FPO
by Melissa M Lawson 04 Aug '08

04 Aug '08

Guilherme Ottoni will present his FPO on Monday August 11 at 10AM in Room 402. The members of his committee are: David August, advisor; Sharad Malik (ELE) and Vivek Sarkar (Rice), readers; Margaret Martonosi (ELE) and David Walker, nonreaders. A copy of his thesis, "Global Instruction Scheduling for Multi-Threaded Architectures", is on display in Room 310. Everyone is invited to attend his talk.

1 0

E Raman preFPO
by Melissa M Lawson 22 May '08

22 May '08

Easwaran Raman will present his preFPO on Thursday May 29, 2PM, in room 402. The members of his committee are: David August, advisor; Teresa Johnson (HP) and Doug Clark, readers; David Walker and Vivek Pai, nonreaders. Everyone is welcome to attend his talk. His abstract follows below. -------------------------------- Title: Extracting Iteration-Level Parallelism from Loops with Inter-Iteration Dependences Abstract --------------- The continuing exponential growth in transistor density and the limitations in using the additional transistors to boost uniprocessor performance has caused processor manufactures to pack multiple processor cores onto a single chip. These processors, known as chip multiprocessors (CMPs) or multicore processors, do not ensure that single-threaded applications continue to run faster unless those applications are parallelized. Automatic parallelization by compilers has a key role to play in utilizing the CMP architecture to improve single-threaded application performance. One of the most common compiler techniques to automatically parallelize loops is DOALL parallelization. DOALL extracts iteration-level parallelism by executing different iterations of the loop concurrently. The advantage of DOALL lies in its scalability with the iteration count of the loop, but its applicability is severely limited by the presence of inter-iteration or loop-carried dependences. My thesis proposes two new techniques to extract iteration-level parallelism from loops with loop-carried dependences. The first technique, known as Speculative Iteration Chunk Execution (Spice), uses value speculation to break loop-carried dependences, enabling speculative execution of chunks of iterations in parallel. Unlike most value-speculation based parallelization techniques, Spice speculates only a few values to unlock parallelism using a software-based value predictor. The second technique is known as Parallel-Stage Decoupled Software Pipelining (PS-DSWP). PS-DSWP extracts both pipelined parallelism and iteration-level parallelism from loops, thereby generalizing two well-known parallelization techniques, DOALL and Decoupled Software Pipelining (DSWP). PS-DSWP partitions a loop into a sequence of pipeline stages that are executed concurrently by different threads, resulting in pipelined parallelism. Iteration-level parallelism is extracted by allowing those stages that do not have any loop-carried dependences, known as parallel stages, to be executed by multiple threads concurrently in a DOALL-like fashion. The applicability and effectiveness of PS-DSWP is further enhanced by using speculation to remove certain dependences. Both these techniques are implemented in the VELOCITY compiler and evaluated on an Itanium 2 CMP simulator. The results demonstrate the ability of these techniques to significantly out-perform prior techniques on several important loops from real applications.

1 0

L Poole general tomorrow - time change
by Melissa M Lawson 21 May '08

21 May '08

Lindsey Poole's general exam scheduled for tomorrow, Thursday May 22, has been moved to 1PM in Room 402.

1 0

L Poole general exam
by Melissa M Lawson 16 May '08

16 May '08

Lindsey Poole will present his research seminar/general exam on Thursday May 22 at 3PM in Room 402. The members of his committee are: Vivek Pai (advisor), Jennifer Rexford, and Larry Peterson. Everyone is invited to attend his talk, and those faculty wishing to remain for the oral exam following are welcome to do so. His abstract and reading list follow below. --------------------------------------------- Abstract While cooperative DNS resolver systems, such as CoDNS, have demonstrated improved reliability and performance over standard approaches, their security has been weaker, since any corruption or misbehavior of a single resolver can easily propagate throughout the system. We address this weakness in a new system called ConfiDNS, which augments the cooperative lookup process with configurable policies that utilize multi-site agreement and per-site lookup histories. Not only does ConfiDNS provide better security than cooperative approaches, but for up to 99.8% of unique lookups, ConfiDNS exceeds the security of standard DNS resolvers. ConfiDNS provides these benefits while retaining the other benefits of CoDNS, such as incremental deployability, higher reliability, and improved performance, in some cases faster than CoDNS. We discuss the implementation of ConfiDNS and the challenges of running a distributed agreement protocol in an environment where DNS-based redirection is is used to facilitate load-balancing and localized content distribution. Textbook [1] L. Peterson, and B. Davie, Computer Networks: A Systems Approach, 3rd edition, Morgan Kaufmann Publishers, May 2003. Research Papers [1] D. Clark, "The design philosophy of the DARPA Internet protocols," ACM SIGCOMM Computer Communication Review, Vol. 18, No. 4, pp. 106-114, Aug. 1988. [2] L. Breslau, P. Cao, L. Fan, G. Phillips, and S. Shenker. Web Caching and Zipf-like Distributions: Evidence and Implications. In 'In Proceedings of IEEE INFOCOM', New York, NY, March 1999. [3] A. Lioy, F. Maino, M. Marian, and D. Mazzocchi. Dns security. In Proceedings of the TERENA Networking Conference, Lisbon, Portugal, May 2000. [4] H. B. Jaeyeon Jung, Emil Sit and R. Morris. DNS Performance and the Effectiveness of Caching. In Proceedings of the ACM SIGCOMM Internet Measurement Workshop '01, San Francisco, California, November 2001. [5] K. Park, V. S. Pai, L. Peterson, and Z. Wang. CoDNS: Improving DNS Performance and Reliability via Cooperative Lookups. In Proceedings of the Sixth Symposium on Operating Systems Design and Implementation (OSDI), San Franscisco, CA, December 2004. [6] P. Mockapetris and K. Dunlap. Development of the Domain Name System. In In Proceedings of the ACM SIGCOMM Conference, Stanford, CA, August 1988. [7] A. Shaikh, R. Tewari, and M. Agrawal. On the effectiveness of dns- based server selection. In Proceedings of INFOCOM 2001. Twentieth Annual Joint Conference of the IEEE Computer and Communications Societies, Anchorage, AK, April 2001. [8] B. Krishnamurthy, C. Wills, and Y. Zhang. On the use and performance of content distribution networks. In Proceedings of SIGCOMM Internet Measurement Workshop, San Francisco, CA, November 2001. [9] H. Ballani and P. Francis. A Simple Approach to DNS DoS Mitigation. In Proceedings of the 5th ACM Workshop on Hot Topics in Networks (HotNets '06), Irvine, CA, November 2006. [10] L. Wang, K. Park, R. Pang, V. Pai, and L. Peterson. Reliability and Security in the CoDeeN Content Distribution Network. In In Proceedings of the USENIX Annual Technical Conference, Boston, MA, June 2004. [11] J. Yin, J. Martin, A. Venkataramani, L. Alvisi, and M. Dahlin. Separating agreement from execution for byzantine fault tolerant services. In In Proceedings of the 19th ACM Symposium on Operating Systems Principles (SOSP-2003), Lake George, NY, October 2003. [12] M. Castro and B. Liskov. Practical byzantine fault tolerance. In In Proceedings of the Symposium on Operating Systems Design and Implementation (OSDI), New Orleans, LA, February 1999.

1 0