December 2017 - talks - lists.cs.princeton.edu

Steven Engelhardt will present his Pre-FPO "Automated discovery of privacy violations on 1 million websites" on January 3, 2018 at 10am in CS 302
by Nicki Gotsis 29 Dec '17

29 Dec '17

Steven Engelhardt will present his Pre-FPO "Automated discovery of privacy violations on 1 million websites" on January 3, 2018 at 10am in CS 302. The members of his committee follow below. Committee members: Arvind Narayanan (adviser), readers: Ed Felten and Nick Feamster; nonreaders: Jen Rexford and Prateek Mittal Everyone is invited to attend his talk. The talk title and abstract follow below. Title: Automated discovery of privacy violations on 1 million websites Abstract: Online tracking is increasingly invasive. Gone are the days where a user can "reset" their online profile by clearing their browser's cookies. Instead, users face persistent, cross-device tracking which blends their offline activity with their online behavior. Tracking protection provided by browsers is often ineffective and circumventable, while solutions based on voluntary cooperation, such as Do Not Track, haven’t had meaningful adoption. Knowledgeable users may turn to anti-tracking tools and ad-blockers for protection, but we find that even these more advanced solutions fail to fully protect against the techniques we study. In this talk, I will present OpenWPM, a platform we developed for flexible and modular web measurement. We’ve used OpenWPM to run repeated, large-scale studies leading to the discovery of numerous privacy and security violations across the web and in emails. These discoveries have helped curtail the adoption of advanced tracking techniques, and have repeatedly informed policy debates and browser privacy decisions. OpenWPM has already been used in over 20 academic studies, including a number of which I’ll highlight in this talk. In particular, I'll discuss the detection of several persistent tracking techniques, including device fingerprinting, cookie syncing, and cookie respawning. We’ve found that nearly every new HTML5 API introduced ends up being misused by trackers for device fingerprinting. The misuse is often invisible to users and publishers alike, and in many cases was not anticipated by API designers. I’ll show how we’re able to use the structure of fingerprinting scripts to detect new techniques, and take a critical look at how the API design process can be changed to prevent such misuse in the future. I'll also explore the budding industry of trackers which use PII-derived identifiers to track users across devices, and even into the offline world. I’ll demonstrate a novel bait technique, which allows us to spoof the presence of PII on a large number of sites. I’ll show how trackers exfiltrate the spoofed PII from websites and emails through the abuse of browser features. Finally, I'll take a critical look at the efforts by browser vendors to protect user privacy and will end with a proposal for a path forward, grounding tracking protection in measurement.

1 0

Victoria Yao will present her Pre FPO "Integrative network-based approaches to analyze genomics data" on Tuesday, December 19, 2017 at 11am in CS 402.
by Nicki Gotsis 15 Dec '17

15 Dec '17

Victoria Yao will present her Pre FPO on Tuesday, December 19, 2017 at 11am in CS 402. The members of her committee are as follows: Adviser: Olga Troyanskaya Readers: Ryan Adams and Mona Singh Non-readers: Barbara Engelhardt and Coleen Murphy (MOL BIO/LSI) The title and abstract for her talk follows below. Title: Integrative network-based approaches to analyze genomics data Abstract: The increasingly commonplace generation of genome-scale data provides us with a wealth of biological knowledge that captures global molecular-level changes in diverse model organisms and humans. However, these large data are often noisy, highly heterogenous, and lack the resolution required to study key aspects of metazoan complexity, such as tissue and cell-type specificity. Furthermore, in order to efficiently apply the results of such studies towards understanding the etiology of complex human diseases, we must consider integrative approaches that fuse computational predictions and model organism experimentation in the context of human data. Here, we apply this core idea to the development of three different network-based approaches, with example applications to the study of neurodegenerative diseases. First, we will discuss diseaseQUEST , an integrative computational-experimental framework that combines human quantitative genetics with in silico functional network representations of model organism biology to systematically identify disease gene candidates. This framework leverages a novel semi-supervised Bayesian network integration approach to predict tissue- and cell-type specific functional relationships between genes in model organisms. We use diseaseQUEST to predict candidate genes for 25 different human diseases and traits using C. elegans as a model system, with a particular focus on Parkinson’s disease. Second, we will describe an approach that takes advantage of high-quality neuron-specific molecular profiles of cells that vary in vulnerability to Alzheimer’s disease obtained in mouse to generate neuron-specific functional networks in human. We then combine these network models with human quantitative genetics data to prioritize likely Alzheimer’s disease candidates. Finally, we will present a network-based approach that systematically identifies differential isoform interactions. We apply this approach to the study of tissue and environment dynamics in Alzheimer’s disease. BQ_BEGIN BQ_END

1 0

4:30pm Thu Dec 14 talk on "Advancing eLearning with data" in E-Quad B205
by Jennifer Rexford 14 Dec '17

14 Dec '17

SEMINAR TODAY: Department of Electrical Engineering Seminar Speaker: Christopher Brinton Zoomi Inc Title: Advancing eLearning with Data Date: Thursday, December 14, 2017 Time: 4:30pm Room: E-Quad, Room B205 Hosts: Profs. Prateek Mittal & H. Vincent Poor Abstract: Electronic learning (eLearning) has witnessed explosive growth over the past decade, with learning scenarios from corporate training to higher education now supporting full course delivery online. Despite the conveniences offered to both learners and instructors, eLearning has been widely criticized for poorer learning outcomes compared with those obtained by traditional classrooms, in terms of metrics such as engagement and knowledge transfer. Today’s learning technology systems, however, can be equipped with infrastructure to capture a plethora of data about learners as they take courses online, such as their multimedia clickstream measurements, assessment responses, and social interactions with one another. The availability of this data presents the opportunity to develop fine-granular models of learners, which can in turn drive methodologies at the intersection of machine learning, data science, and the science of learning aimed at improving learning outcomes. In this talk, I will describe the research I have undertaken to improve the quality of eLearning, while at my company Zoomi Inc., during my graduate studies, and as an instructor. Broadly speaking, my team’s investigations have followed three, complementary thrusts: Predictive Learning Analytics (PLA), Social Learning Networks (SLN), and Deep Learning Personalization (DLP). PLA involves developing algorithms for early detection of learning outcomes that model from behavioral data and can provide actionable analytics to instructors. SLN involves quantifying how learners seek and disseminate knowledge socially, and designing algorithms to predict and optimize these processes. DLP involves systematically dividing course content into micro-segments, and combining topic, learner, and reinforcement learning models to adapt the material presented to individual learners in real time. In describing these thrusts, I will point out results Zoomi has witnessed in deployments to different learning scenarios, and will also discuss open research questions. Bio: Dr. Christopher G. Brinton is the Head of Advanced Research at Zoomi Inc, an AI-driven learning technology company he co-founded in 2013, and a Lecturer in Electrical Engineering at Princeton University. His research focus is developing systems and methods to improve the quality of student learning, through predictive learning analytics, social learning networks, and individualization. Chris co-authored the book The Power of Networks: Six Principles that Connect our Lives, and has reached over 250,000 students through MOOCs based on his book. A recipient of the 2016 Bede Liu Best Dissertation Award in Electrical Engineering, Chris received his PhD from Princeton in 2016, his Master’s from Princeton in 2013, and his BSEE from The College of New Jersey (valedictorian and summa cum laude) in 2011, all in Electrical Engineering.

1 0

Haoyu Zhang will present his Pre FPO talk "Resource Management for Advanced Data Analytics at Large Scale" on Tuesday December 19th, 2017 at 10:00am in CS 302.
by Nicki Gotsis 12 Dec '17

12 Dec '17

Haoyu Zhang will present his Pre FPO talk on Tuesday December 19th, 2017 at 10:00am in CS 302. The members of his committee are: Michael Freedman (advisor), Kyle Jamieson, Wyatt Lloyd (readers), Kai Li, Jennifer Rexford (examiners). All are welcome to attend. Title: Resource Management for Advanced Data Analytics at Large Scale The rapidly growing size of data and the complexity of analytics present new challenges for large-scale data processing systems. Modern distributed computing frameworks need to support not only embarrassingly parallelizable batch jobs, but also advanced applications analyzing text and multimedia data using multi-stage DAG queries and machine learning (ML) models. Given the high costs of advanced data analytics, resource management is crucial. New applications and workloads expose vastly different characteristics which makes traditional scheduling systems inadequate, and at the same time offer many opportunities that lead to new system designs for better performance. In this talk, I will present resource management system designs that can efficiently utilize cluster resources by leveraging the insights from advanced data analytics applications. We identify and study the following three key scenarios: (i) VideoStorm: a video analytics system that processes thousands of video analytics queries on live video streams over large clusters. VideoStorm's offline scheduler generates resource-quality profiles for vision processing queries, and its online scheduler allocates resources to maximize performance in terms of quality and lag, in contrast to the commonly used fair sharing of resources. (ii) SLAQ: a cluster scheduling system for approximate ML training jobs that aims to maximize the overall job quality. ML training typically involves many iterations that continually improve the quality of the model. Yet in exploratory settings, better models can be obtained faster by directing resources to jobs with the most potential for improvement. SLAQ collects quality and resource usage information from concurrent jobs, and allocates resources to maximize quality improvement based on highly-tailored model quality predictions. (iii) Riffle: an optimized shuffle service for big-data analytics frameworks that significantly improves I/O efficiency and scales to process petabytes of data. The all-to-all data transfer (i.e., shuffle) in modern big-data systems (such as Spark and Hadoop) becomes the scaling bottleneck for multi-stage analytics jobs, due to the superlinear increase in disk I/O operations as data volume increases. Riffle significantly improves system performance by merging fragmented intermediate files and efficiently scheduling the merge operations. We have built and deployed the systems at large clusters and performed extensive evaluation with real production workloads. Our results show significant improvement in resource efficiency, job completion time, and system throughput.

1 0

Colloquium Speaker Duncan Watts TODAY at 12:30
by Emily Lawrence 12 Dec '17

12 Dec '17

Colloquium Speaker Duncan Watts, Microsoft Research Tuesday, December 12, 2017 - 12:30pm Computer Science - Room 105 Hosts: Jennifer Rexford / Mitchell Duneier (Sociology) Event page: http://www.cs.princeton.edu/events/25605 Title: "Computational Social Science: Exciting Progress and Future Challenges" "The past 15 years have witnessed a remarkable increase in both the scale and scope of social and behavioral data available to researchers, leading some to herald the emergence of a new field: "computational social science." In this talk I highlight two areas of research that would not have been possible just a handful of years ago: first, using "big data" to study social contagion on networks; and second, using virtual labs to extend the scale, duration, and complexity of traditional lab experiments. Although these examples were all motivated by substantive problems of longstanding interest to social science, they also illustrate how new classes of data can cast these problems in new light. At the same, they illustrate some important limitations faced by our existing data generating platforms. I then conclude with some thoughts on how CSS might overcome some of these obstacles to progress." Bio: Duncan Watts is a principal researcher at Microsoft Research and a founding member of the MSR-NYC lab. He is also an AD White Professor at Large at Cornell University. Prior to joining MSR in 2012, he was from 2000-2007 a professor of Sociology at Columbia University, and then a principal research scientist at Yahoo! Research, where he directed the Human Social Dynamics group. His research on social networks and collective dynamics has appeared in a wide range of journals, from Nature, Science, and Physical Review Letters to the American Journal of Sociology and Harvard Business Review, and has been recognized by the 2009 German Physical Society Young Scientist Award for Socio and Econophysics, the 2013 Lagrange-CRT Foundation Prize for Complexity Science, and the 2014 Everett Rogers M. Rogers Award. He is also the author of three books: Six Degrees: The Science of a Connected Age (W.W. Norton, 2003) and Small Worlds: The Dynamics of Networks between Order and Randomness (Princeton University Press, 1999), and most recently Everything is Obvious: Once You Know The Answer (Crown Business, 2011). Watts holds a B.Sc. in Physics from the Australian Defense Force Academy, from which he also received his officer's commission in the Royal Australian Navy, and a Ph.D. in Theoretical and Applied Mechanics from Cornell University.

1 0

Yann LeCun public lecture today at the IAS
by Sanjeev Arora 12 Dec '17

12 Dec '17

Yann LeCun, director of Facebook AI Research, is giving a popular talk at the IAS today (Tuesday) at 5:30pm. https://www.ias.edu/events/lecun-publiclecture (Sorry, I realize it conflicts with the CS holiday party. It was planned quite a while ago.) Best, Sanjeev

1 0

Reminder: Colloquium Speaker Duncan Watts, tomorrow at 12:30pm
by Emily Lawrence 11 Dec '17

11 Dec '17

Colloquium Speaker Duncan Watts, Microsoft Research Tuesday, December 12, 2017 - 12:30pm Computer Science - Room 105 Hosts: Jennifer Rexford / Mitchell Duneier (Sociology) Event page: http://www.cs.princeton.edu/events/25605 Title: "Computational Social Science: Exciting Progress and Future Challenges" The past 15 years have witnessed a remarkable increase in both the scale and scope of social and behavioral data available to researchers, leading some to herald the emergence of a new field: "computational social science." In this talk I highlight two areas of research that would not have been possible just a handful of years ago: first, using "big data" to study social contagion on networks; and second, using virtual labs to extend the scale, duration, and complexity of traditional lab experiments. Although these examples were all motivated by substantive problems of longstanding interest to social science, they also illustrate how new classes of data can cast these problems in new light. At the same, they illustrate some important limitations faced by our existing data generating platforms. I then conclude with some thoughts on how CSS might overcome some of these obstacles to progress. Bio: Duncan Watts is a principal researcher at Microsoft Research and a founding member of the MSR-NYC lab. He is also an AD White Professor at Large at Cornell University. Prior to joining MSR in 2012, he was from 2000-2007 a professor of Sociology at Columbia University, and then a principal research scientist at Yahoo! Research, where he directed the Human Social Dynamics group. His research on social networks and collective dynamics has appeared in a wide range of journals, from Nature, Science, and Physical Review Letters to the American Journal of Sociology and Harvard Business Review, and has been recognized by the 2009 German Physical Society Young Scientist Award for Socio and Econophysics, the 2013 Lagrange-CRT Foundation Prize for Complexity Science, and the 2014 Everett Rogers M. Rogers Award. He is also the author of three books: Six Degrees: The Science of a Connected Age (W.W. Norton, 2003) and Small Worlds: The Dynamics of Networks between Order and Randomness (Princeton University Press, 1999), and most recently Everything is Obvious: Once You Know The Answer (Crown Business, 2011). Watts holds a B.Sc. in Physics from the Australian Defense Force Academy, from which he also received his officer's commission in the Royal Australian Navy, and a Ph.D. in Theoretical and Applied Mechanics from Cornell University.

1 0

Max Homilius will present his Pre FPO, "Prioritizing animal models and drug targets using functional networks", on Monday, December 11th, 2017 at 12 pm in LSI 260.
by Nicki Gotsis 06 Dec '17

06 Dec '17

Max Homilius will present his Pre FPO, "Prioritizing animal models and drug targets using functional networks", on Monday, December 11th, 2017 at 12 pm in LSI 260. The members of his committee are as follows: Advisor: Olga Troyanskaya Readers: Mona Singh, Tilo Grosser (UPenn) Non-readers: Barbara Engelhardt, Coleen Murphy (Mol Bio) Abstract and title follows below. Title: Prioritizing animal models and drug targets using functional networks Abstract: Living cells maintain a complex network of biological interactions, including metabolic reactions, physical protein-protein interactions and the regulation of gene expression. Representing these interactions as complex networks can reveal molecular mechanisms underlying human disease and adverse events due to drug treatment. In my thesis I make use of functional interaction networks, which effectively predict gene function from noisy, incomplete, and heterogeneous genomics data, to develop computational approaches for human disease research and to gain insight in the mode of action of drugs. First, I will speak about the inference of tissue- and process-specific functional interaction networks tailored to study the biology non-steroidal anti-inflammatory drugs (NSAIDs). Next, I will present a computational approach for prioritizing animal models for human diseases. Apart from approaches reliant on shared homologous genes, we lack tools to seamlessly search across organisms to identify the model phenotype equivalent to a human disease. By jointly using genome-scale functional interaction networks in both human and model organism, I create and match genome-wide representations of human diseases and model phenotypes, and report novel genes which are prime candidates for experimental follow-up. My third project investigates the cellular mode-of-action of drugs, which is critical in assessing both beneficial and adverse effects of treatment. Transcriptional profiling is a powerful approach towards this goal and has been used extensively to identify genes and pathways that respond to drug treatment. However, the concentration required to elicit measurable changes in gene-expression is often significantly higher than a drug’s therapeutic dose. Therefore, to gain clinically relevant insights, I developed a network-based method that can detect critical genes and pathways even in the absence of large-scale expression differences in response to drug treatment.

1 0

Fwd: EE Seminar - December 14, 2017, 4:30pm, E-Quad B205 - Christopher Brinton, Zoomi Inc
by Jennifer Rexford 05 Dec '17

05 Dec '17

> From: "Lisa R. Lewis" <ll2(a)Princeton.EDU> > Subject: EE Seminar - December 14, 2017, 4:30pm, E-Quad B205 - Christopher Brinton, Zoomi Inc > Date: December 5, 2017 at 3:36:16 PM EST > To: ee-seminar(a)Princeton.EDU > Reply-To: "Lisa R. Lewis" <ll2(a)Princeton.EDU> > > > Department of Electrical Engineering > Seminar > > Speaker: Christopher Brinton > Zoomi Inc > Title: Advancing eLearning with Data > Date: Thursday, December 14, 2017 > Time: 4:30pm > Room: E-Quad, Room B205 > Hosts: Profs. Prateek Mittal & H. Vincent Poor > > Abstract: Electronic learning (eLearning) has witnessed explosive growth over the past decade, with learning scenarios from corporate training to higher education now supporting full course delivery online. Despite the conveniences offered to both learners and instructors, eLearning has been widely criticized for poorer learning outcomes compared with those obtained by traditional classrooms, in terms of metrics such as engagement and knowledge transfer. Today’s learning technology systems, however, can be equipped with infrastructure to capture a plethora of data about learners as they take courses online, such as their multimedia clickstream measurements, assessment responses, and social interactions with one another. The availability of this data presents the opportunity to develop fine-granular models of learners, which can in turn drive methodologies at the intersection of machine learning, data science, and the science of learning aimed at improving learning outcomes. > > In this talk, I will describe the research I have undertaken to improve the quality of eLearning, while at my company Zoomi Inc., during my graduate studies, and as an instructor. Broadly speaking, my team’s investigations have followed three, complementary thrusts: Predictive Learning Analytics (PLA), Social Learning Networks (SLN), and Deep Learning Personalization (DLP). PLA involves developing algorithms for early detection of learning outcomes that model from behavioral data and can provide actionable analytics to instructors. SLN involves quantifying how learners seek and disseminate knowledge socially, and designing algorithms to predict and optimize these processes. DLP involves systematically dividing course content into micro-segments, and combining topic, learner, and reinforcement learning models to adapt the material presented to individual learners in real time. In describing these thrusts, I will point out results Zoomi has witnessed in deployments to different learning scenarios, and will also discuss open research questions. > > Bio: Dr. Christopher G. Brinton is the Head of Advanced Research at Zoomi Inc, an AI-driven learning technology company he co-founded in 2013, and a Lecturer in Electrical Engineering at Princeton University. His research focus is developing systems and methods to improve the quality of student learning, through predictive learning analytics, social learning networks, and individualization. Chris co-authored the book The Power of Networks: Six Principles that Connect our Lives, and has reached over 250,000 students through MOOCs based on his book. A recipient of the 2016 Bede Liu Best Dissertation Award in Electrical Engineering, Chris received his PhD from Princeton in 2016, his Master’s from Princeton in 2013, and his BSEE from The College of New Jersey (valedictorian and summa cum laude) in 2011, all in Electrical Engineering.

1 0

Colloquium Speaker Duncan Watts, Dec 12
by Emily Lawrence 05 Dec '17

05 Dec '17

Colloquium Speaker Duncan Watts, Microsoft Research Tuesday, December 12, 2017 - 12:30pm Computer Science - Room 105 Hosts: Jennifer Rexford / Mitchell Duneier (Sociology) Title: "Computational Social Science: Exciting Progress and Future Challenges" Abstract: The past 15 years have witnessed a remarkable increase in both the scale and scope of social and behavioral data available to researchers, leading some to herald the emergence of a new field: "computational social science." In this talk I highlight two areas of research that would not have been possible just a handful of years ago: first, using "big data" to study social contagion on networks; and second, using virtual labs to extend the scale, duration, and complexity of traditional lab experiments. Although these examples were all motivated by substantive problems of longstanding interest to social science, they also illustrate how new classes of data can cast these problems in new light. At the same, they illustrate some important limitations faced by our existing data generating platforms. I then conclude with some thoughts on how CSS might overcome some of these obstacles to progress. Bio: Duncan Watts is a principal researcher at Microsoft Research and a founding member of the MSR-NYC lab. He is also an AD White Professor at Large at Cornell University. Prior to joining MSR in 2012, he was from 2000-2007 a professor of Sociology at Columbia University, and then a principal research scientist at Yahoo! Research, where he directed the Human Social Dynamics group. His research on social networks and collective dynamics has appeared in a wide range of journals, from Nature, Science, and Physical Review Letters to the American Journal of Sociology and Harvard Business Review, and has been recognized by the 2009 German Physical Society Young Scientist Award for Socio and Econophysics, the 2013 Lagrange-CRT Foundation Prize for Complexity Science, and the 2014 Everett Rogers M. Rogers Award. He is also the author of three books: Six Degrees: The Science of a Connected Age (W.W. Norton, 2003) and Small Worlds: The Dynamics of Networks between Order and Randomness (Princeton University Press, 1999), and most recently Everything is Obvious: Once You Know The Answer (Crown Business, 2011). Watts holds a B.Sc. in Physics from the Australian Defense Force Academy, from which he also received his officer's commission in the Royal Australian Navy, and a Ph.D. in Theoretical and Applied Mechanics from Cornell University.

1 0