May 2013 - talks - lists.cs.princeton.edu

Starting Now: Udi Weinsberg Talk, Sherrerd 306: Building a Practical Privacy-Preserving Recommender System
by Joshua A. Kroll 29 May '13

29 May '13

Udi Weinsberg of Technicolor will be speaking at 2:00pm on May 29, in Sherrerd 306. His title and abstract are below. Building a Practical Privacy-Preserving Recommender System Many online services, such as recommender systems, email, and social networks collect user data, which is then used for both personalization and monetization. Although the latter enables services to be free, users are realizing that these services come at a hidden cost of potentially exposing their private data. In this talk I will show that even the common 5-star item-rating recommender system leaks private demographic information. Then, I will discuss methods for helping users preserve their privacy while getting accurate recommendations. Finally, a building block of many recommender systems, and an important machine-learning algorithm on its own, is linear regression. I will present a system that learns a linear model without learning anything about the private input data other than the output model.

1 0

Two Talks on Privacy: Tuesday 5/28 and Wednesday 5/29
by Joshua A. Kroll 27 May '13

27 May '13

Hi Everyone, The Security Group is holding two great talks on new research in privacy. The first is Tuesday, 5/28 at 4pm and the second is Wednesday, 5/29 at 2pm. Both are in Sherrerd Hall 306 (the CITP conference room). Please see the notices below. Thanks, Josh -------------------- Matthew Wright of the University of Texas at Arlington will be speaking at 4pm on May 28th (tomorrow!) in Sherrerd 306. His title and abstract are below. Leveraging Social Networks for Improved Anonymity and P2P Systems Social networks are great for connecting with other people, but they can also be leveraged for enhanced security properties. In this talk, I will describe two systems -- Pisces and Persea -- that we have designed to take advantage of the information that is inherent in the social network structure. Pisces is a system for enhancing anonymity in peer-to-peer (P2P) anonymity system designs. An anonymity system, such as the popular Tor network, helps protect your privacy on the Internet and enables people in countries like Syria to get around Internet censorship. In Pisces, we route our anonymity paths through users' social connections using verifiable random paths. We show that this technique provides much better privacy than prior designs in the face of strong attackers. Persea addresses the reliability of looking up information and resources in a P2P system, such as Skype or Bittorrent. Existing systems are vulnerable to an attacker adding many malicious peer nodes, e.g. by using a botnet, and having them undermine the reliability of lookups. We propose a P2P system, Persea, based on a bootstrap tree -- essentially a social network that shows how each person entered the P2P system via a series of invitations. We embed the bootstrap tree into the identities that nodes use to locate themselves and perform lookups. We argue that this approach is more suitable to P2P systems than prior approaches and show that it provides lookup success rates at least as good as in prior work. -------------------- Udi Weinsberg of Technicolor will be speaking at 2:00pm on May 29, in Sherrerd 306. His title and abstract are below. Building a Practical Privacy-Preserving Recommender System Many online services, such as recommender systems, email, and social networks collect user data, which is then used for both personalization and monetization. Although the latter enables services to be free, users are realizing that these services come at a hidden cost of potentially exposing their private data. In this talk I will show that even the common 5-star item-rating recommender system leaks private demographic information. Then, I will discuss methods for helping users preserve their privacy while getting accurate recommendations. Finally, a building block of many recommender systems, and an important machine-learning algorithm on its own, is linear regression. I will present a system that learns a linear model without learning anything about the private input data other than the output model.

1 0

Jialu Huang FPO
by Melissa M. Lawson 24 May '13

24 May '13

Jialu Huang will present her FPO on Monday June 3 at Noon in room 402. The members of her committee are: David August, advisor; Kai Li and David Walker, readers; Doug Clark and JP Singh, nonreaders. A copy of her thesis, "Automatically Exploiting Cross-Invocation Parallelism Using Runtime Information", is on display in Room 310. Everyone is invited to attend her talk.

1 0

Fwd: Mehmet E Basbug General Exam
by Melissa M. Lawson 20 May '13

20 May '13

Mehmet is an ELE student who is advised by Rob Schapire. He's asked that we announce his generals in case anyone from CS is interested in attending. ELE generals do not post a reading list like CS does. ---------------------------------------------------------------------- Mehmet E Basbug will present his research seminar/general exam on Monday May 20 at 1:30pm in E-Quad Room J323. The members of his committee are: Robert Schapire (advisor), Kai Li , and Peter Ramadge. 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 is given below: Abstract ======== The decoding of brain states using machine-learning and pattern-classification techniques has become a widely used approach in many areas of neuroscience research. Decoding methods are especially essential to the development of non-invasive brain-computer interfaces. One of the most exciting prospects of decoding in basic neuroscience is the potential to modify a person's sensory cortex in response to brain events as they happen or perhaps slightly before they happen. We used EEG and MEG signals to decode the intention of movement even before the movement happens. In doing that we took a big data approach to the problem and developed reliable and scalable models for boosting algorithm.

1 0

N Katta generals
by Melissa M. Lawson 17 May '13

17 May '13

Naga Katta will present his research seminar/general exam on Wednesday May 22 at 2PM in Room 402. The members of his committee are: Jennifer Rexford (advisor), David Walker, and Michael Freedman. 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: The \emph{network update problem} requires that a network operator replace the current, global network configuration with a new, global network configuration while preserving key correctness invariants. One way to solve the problem is by ensuring \emph{per-packet} or \emph{per-flow consistency}. The former ensures that every packet traverses either the old network configuration, or the new configuration, but not some combination of the two. The latter generalizes the former by ensuring in addition that every packet in the same flow traverses the same configuration. A consequence of either consistency model is that if both old and new configuration share some property of their paths through the network, such as loop-freedom, access control or connectivity then that property is guaranteed to persist across an update. A disadvantage of past implementation mechanisms for these consistency models is that the transition from one configuration to the next can double the configuration rule-space requirements on a network switch. we present general-purpose algorithms for solving the network update problem, that are designed to trade the time required to perform an update against the additional rule-space needed. More specifically, we break a global consistent update in to $K$ rounds, with each round transferring some subset of the network traffic from old to new configuration. The more rounds used, the longer the update time, but the less rule space overhead is required. To ensure consistency, we perform an analysis of old and new configurations, determining the rules that must be added and may be removed from a switch in each round. To determine the optimal use of rule space, we demonstrate how to represent the optimization problem as a mixed integer linear program. We also show how to extend the mixed integer linear program to additionally optimize for rapid update of the largest flows. Finally, we present initial empirical results that illustrate the tradeoff between time and space. Reading List: Book [1] L. Peterson, and B. Davie, "Computer Networks: A Systems Approach", Morgan Kaufmann, 5e, 2011 Research Papers [2] D. Clark, "The Design Philosophy of the DARPA Internet Protocols", ACM SIGCOMM, August 1988 [3] J.Saltzer, D.Reed and D.Clark, "End-to-End Arguments in System Design", ACM Transactions on Computer Systems, Nov 1984 [4] Ethane: Taking Control of the Enterprise, Martìn Casado, Michael J. Freedman, Justin Pettit, Jianying Luo, Nick McKeown, Scott Shenker, Sigcomm 2007 [5] Onix: A Distributed Control Platform for Large-scale Production Networks, Teemu Koponen, Martìn Casado, Natasha Gude, Jeremy Stribling, Leon Poutievski, Min Zhu, Rajiv Ramanathan, Yuichiro Iwata, Hiroaki Inoue, Takayuki Hama, Scott Shenker, OSDI 2010 [6] Abstractions for Network Update, Mark Reitblatt, Nate Foster, Jennifer Rexford, Cole Schlesinger, David Walker, Sigcomm 2012 [7] Header space analysis: static checking for networks. Peyman Kazemian, George Varghese, and Nick McKeown, (NSDI'12). [8] A Safe, Efﬁcient Update Protocol for OpenFlow Networks, Rick McGeer, HotSDN 2012 [9] DevoFlow: Scaling Flow Management for High-performance Networks, Andrew R. Curtis, Jeffrey C. Mogul, Jean Tourrilhes, Praveen Yalagandula, Puneet Sharma, Sujata Banerjee, Sigcomm 2011 [10] TCAM Razor: a systematic approach towards minimizing packet classifiers in TCAMs, Alex X. Liu, Chad R. Meiners, and Eric Torng, IEEE/ACM Transactions on Networking (TON) [11] Using CPU as a Traffic Co-processing Unit in Commodity Switches, Guohan Lu, Rui Miao, Yongqiang Xiong, Chuanxiong Guo, HotSDN 2012 [12] OFLOPS: An Open Framework for OpenFlow Switch Evaluation, Charalampos Rotsos, Nadi Sarrar, Steve Uhlig, Rob Sherwood, Andrew W. Moore, PAM 2012 [13] Flyways To Decongest Data Center Networks., S. Kandula, J. Padhye, and P. Bahl. In Proc. of HotNets, 2005.

1 0

P Przytycki generals
by Melissa M. Lawson 17 May '13

17 May '13

Pawel Przytycki will present his research seminar/general exam on Wednesday May 22 at 10AM in Room 402. The members of his committee are: Mona Singh (advisor), Olga Troyanskaya, and Andrea LaPaugh. Everyone is welcome 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 Even though the first human genome was sequenced more than a decade ago, it is largely still unknown what specific genetic variations contribute to the diversity of human life. It is only recently that whole-genome sequencing has become affordable enough to utilize at a large enough scale to precisely examine the genetic differences between people. Furthermore, exome sequencing, which is sequencing only the 2% of the human genome that codes for proteins, has surfaced as a cheap way to find minute differences between people’s genetic codes. I propose a method called Nexome designed for network driven analysis of exome sequencing data. Nexome is a degree-aware gene prioritization algorithm that uses a random walker method to diffuse variation information mapped to genes over a protein-protein interaction network. As opposed to many previous algorithms which rely on a pre-defined set of seed genes, Nexome is designed to work with dense non-binary data. Degree correction is derived from diffusion of an uniformed prior. Nexome confirms that there is modularity in cancer variation data when it is mapped to the protein interaction network. Furthermore, this method is capable of discovering known cancer genes that are not frequently mutated. Finally, when applied to specific cancer types Nexome reveals that there is a modular structure to genes that are uniquely enriched in these cancers. BOOK An Introduction to Bioinformatics Algorithms - Pevzner and Jones PAPERS Erten, S., Bebek, G., Ewing, R. M., & Koyutürk, M. (2011). DADA: Degree-Aware Algorithms for Network-Based Disease Gene Prioritization. BioData mining, 4(1), 19. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/21699738 Fu, W. et. al. (2013). Analysis of 6,515 exomes reveals the recent origin of most human protein-coding variants. Nature 493, 216–220. Retreived from: http://www.nature.com/nature/journal/v493/n7431/full/nature11690.html Kim, Y., Przytycka, A. (2012). Bridging the Gap between Genotype and Phenotype via Network Approaches. Frontiers in Genetics, 3, 00227. Retrieved from http://www.frontiersin.org/Journal/Abstract.aspx?s=1187&name=statistical_ge… Lee, I., Blom, U. M., Wang, P. I., Shim, J. E., & Marcotte, E. M. (2011). Prioritizing candidate disease genes by network-based boosting of genome-wide association data. Genome Research, 21(7), 1109–1121. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/21536720 Navlakha, S., & Kingsford, C. (2010). The power of protein interaction networks for associating genes with diseases. Bioinformatics, 26(8), 1057–1063. Retrieved from http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2853684&tool=pmce… O’Roak, B. J., Vives, L., Girirajan, S., Karakoc, E., Krumm, N., Coe, B. P., Levy, R., et al. (2012). Sporadic autism exomes reveal a highly interconnected protein network of de novo mutations. Nature, 485(7397), 1-7. Nature Publishing Group. Retrieved from http://www.nature.com/doifinder/10.1038/nature10989 The Cancer Genome Atlas Research Network. (2012). Comprehensive genomic characterization of squamous cell lung cancers. Nature 489, 519–525. Retreived from: http://www.nature.com/nature/journal/v489/n7417/full/nature11404.html Rossin, E.J. (2011). Proteins Encoded in Genomic Regions Associated with Immune-Mediated Disease Physically Interact and Suggest Underlying Biology. Plos Genetics 7(1). Retrieved from http://www.plosgenetics.org/article/info%3Adoi%2F10.1371%2Fjournal.pgen.100… Wang, P. I., & Marcotte, E. M. (2010). It’s the machine that matters: Predicting gene function and phenotype from protein networks. Journal of proteomics, 73(11), 2277–2289. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/20637909 1000 Genomes Project Consortium. (2010). A map of human genome variation from population-scale sequencing. Nature 467, 1061–1073. Retreived from: http://www.nature.com/nature/journal/v467/n7319/full/nature09534.html

1 0

12-1pm Mon May 20 talk on ICT convergence by CTO of KT (Korea Telecom) in E-Quad B327
by Jennifer Rexford 17 May '13

17 May '13

Speaker: Dr. James Won-Ki Hong, CTO KT (Korea Telecom) Title: Shaping the Future Society with ICT Convergence Date: Monday, May 20, 2013 Time: 12:00 pm Room: E-Quad, Room B327 Host: Princeton EDGE Lab and Princeton KSA Abstract: Korea, also known as the nation of Psy, is one of the most dynamic countries in the world. Thanks to advanced information and communications infrastructures and unique culture, people in Korea are said to be living in the future of many other countries. It is this customer environment that facilitated the emergence of instant messaging and social platforms such as Kakaotalk and Line. It may be surprising to know that the entire country of Korea was actually devastated by the Korean War more than 60 years ago. Many factors drove Korea’s development for the past 60 years. Not only did its high education fever and speedy culture but also the government's strong will and drive on ICT (Information and Communications Technology) development and decisive investments by private sector businesses, such as KT (formerly Korea Telecom), contributed to the Korea’s development. In fact, KT led the deployment of the most advanced wireline/wireless Internet infrastructure in the world during the last few decades. Now, KT is continuing to shape the society we live in with further advances in ICT. To bring the future into reality, KT is taking on numerous challenging and interesting problems in ICT convergence areas. In this talk, we will provide an outline of Korea’s ICT status, KT’s efforts in infrastructure deployment, and peek into future lifestyles based on emerging technologies. We also offer opportunities for talented students and researchers to collaborate with KT.

1 0

R Ranganath generals - time change to 10:30
by Melissa M. Lawson 16 May '13

16 May '13

The start time for Rajesh's generals on Tuesday May 21 has been moved to 10:30. Please note. thanks, Melissa

1 0

Hanjun Kim FPO
by Melissa M. Lawson 16 May '13

16 May '13

Hanjun Kim will present his FPO on Wednesday May 22 at 9:30 AM in Room 302 (note room!) The members of his committee are: David August, advisor; Kai Li and David Wentzlaff (ELE), readers; JP Singh and Sharad Malik (ELE), nonreaders. A copy of his thesis, "ASAP: Automatic Speculative Acyclic Parallelization for Clusters", is on display in Room 310. Everyone is invited to attend his talk.

1 0

R Ranganath generals
by Melissa M. Lawson 16 May '13

16 May '13

Rajesh Ranganath will present his research seminar/general exam on Tuesday May 21 at 10AM in Room 402. The members of his committee are: David Blei (advisor), Rob Schapire, and Philippe Rigollet (ORF). 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. ---------------------- Title: An Adaptive Learning Rate for Stochastic Variational Inference Abstract: Stochastic variational inference finds good pos- terior approximations of probabilistic models with very large data sets. It optimizes the variational objective with stochastic optimization, following noisy estimates of the natural gradient. Operationally, stochastic inference iteratively subsamples from the data, analyzes the subsample, and updates parameters with a decreasing learning rate. However, the algorithm is sensitive to that rate, which usually requires hand-tuning to each application. We solve this problem by developing an adaptive learning rate for stochastic inference. Our method requires no tuning and is easily implemented with computations already made in the algorithm. We demonstrate our approach with latent Dirichlet allocation applied to three large text corpora. Inference with the adaptive learning rate converges faster and to a better approximation than the best settings of hand-tuned rates. This is joint work with Chong Wang, Dave Blei, and Eric Xing. Reading List: Textbook Chapters Pattern Recognition & Machine Learning, Chapters 8, 10, 11 (C. Bishop) Probability & Stochastics, Chapter 6 (E. Cinlar.) Artificial Intelligence: A Modern Approach 3rd edition, Chapters 17, 18, 20, 21 (Russel + Norvig) Papers: Integration of Early Physiological Responses Predicts Later Illness Severity in Preterm Infants [Sci Transl Med 8 September 2010] (S. Saria et al.) Learning individual and population level traits from Clinical Temporal data [NIPS 2010] (S. Saria et al.) {Posterior Predictive Assessment of Model Fitness Via Realized Discrepancies [Statistica Sinica 1996] (A. Gelman, X. Meng, H. Stern) Stochastic Variation Inference [JMLR to appear] (M. Hoffman et al) A Stochastic Gradient Method with an Exponential Convergence Rate for Strongly-Convex Optimization with Finite Training Sets. [NIPS 2012] (N. LeRoux et al) Bootstrap Methods for Standard Errors, Confidence Intervals, and Other Measures of Statistical Accuracy [Statistical Science 1986] (B. Efron, R. Tibishirani) The Indian Buffet Process: An Introduction and Review [JMLR 2011] (T. Griffiths, Z. Gharamani)

1 0