Artificial Intelligence Seminar

Spring 2011
Friday 12:00-1:15
Upson 5130

The AI seminar will meet weekly for lectures by graduate students, faculty, and researchers emphasizing work-in-progress and recent results in AI research. Lunch will be served starting at noon, with the talks running between 12:15 and 1:15. The new format is designed to allow AI chit-chat before the talks begin. Also, we're trying to make some of the presentations less formal so that students and faculty will feel comfortable using the seminar to give presentations about work in progress or practice talks for conferences.
February 4th

Speaker: Thorsten Joachims (or a student of Thorsten's)

Title : Dynamic Ranked Retrieval

Abstract :  Most queries issued to a search engine are ambiguous at some level. This presents the search engine with a dilemma. On the one hand, if it focuses on the most likely interpretation of the query, it does not provide any utility to users that had a different intent. On the other hand, if it tries to provides at least one relevant results for all query intents, it will not cover any intent particularly well.

We present a retrieval model that combines these otherwise contradictory goals of result diversification and high coverage. The key idea is to make the result ranking dynamic, allowing limited and well-controllable change based on interactive user feedback. Unlike conventional rankings that remain static after the query was issued, a dynamic ranking allows and anticipates user activity. We develop a decision-theoretic framework to guide the design and evaluation of algorithms for this interactive retrieval setting. Furthermore, we propose two dynamic ranking algorithms, both of which are computationally efficient. We prove that these algorithms provide retrieval performance that is guaranteed to be at least as good as the optimal static ranking algorithm. In empirical evaluations, dynamic ranking shows substantial improvements in retrieval performance over conventional static rankings.

Joint work of Christina Brandt, Thorsten Joachims, Yisong Yue, and Jacob Bank.

“The AI-Seminar is sponsored by Yahoo!”

February 11th

Speaker: Jim Rehg

Host: Dan Huttenlocher

Bio: James M. Rehg (pronounced "ray") is a Professor in the School of Interactive Computing at the Georgia Institute of Technology, where he is the Director of the Center for Behavior Imaging, co-Director of the Computational Perception Lab, and Associate Director of Research in the Center for Robotics and Intelligent Machines. He received his Ph.D. from CMU in 1995 and worked at the Cambridge Research Lab of DEC (and then Compaq) from 1995-2001, where he managed the computer vision research group. He received the National Science Foundation (NSF) CAREER award in 2001, and the Raytheon Faculty Fellowship from Georgia Tech in 2005. He and his students have received a number of best paper awards, including best student paper awards at ICML 2005 and BMVC 2010. Dr. Rehg is active in the organizing committees of the major conferences in computer vision, most-recently serving as the General co-Chair for IEEE CVPR 2009. He has served on the Editorial Board of the International Journal of Computer Vision since 2004. He has authored more than 100 peer-reviewed scientific papers and holds 23 issued US patents. Dr. Rehg is currently leading a multi-institution effort to develop the science and technology of Behavior Imaging, funded by an NSF Expedition award (see for details).

Title :  Temporal Causality and the Analysis of Interactions in Video

Abstract :  A basic goal of video understanding is the organization of video data into sets of events with associated temporal dependencies. For example, a soccer goal could be explained using a vocabulary of events such as passing, dribbling, tackling, etc. In describing the dependencies between events it is natural to invoke the concept of causality, but previous attempts to perform causal reasoning in video analysis have been limited to special cases, such as sporting events or naïve physics, where strong domain models are available. In this talk I will describe a novel, data-driven approach to the analysis of causality in video. The key to our approach is the representation of low-level visual events as the output of a multivariate point process, and the use of a nonparametric formulation of temporal causality to group event data into interacting subsets. This grouping process differs from standard motion segmentation methods in that it exploits the temporal structure in video over extended time scales. Our method is particularly well-suited to the analysis of social interactions, as it provides a means to organize sensor data and expose patterns of back-and-forth interaction. I will present results for categorizing and retrieving social games between parents and children from unstructured video collections. This application is part of a larger effort in using sensing, machine learning, and AI technologies to support the detection, treatment, and understanding of developmental disorders such as autism. I will present a brief overview of these activities, which are supported by a 2010 Expeditions in Computing Award from the National Science Foundation.
This is joint work with Karthir Prabhakar, Sangmin Oh, Ping Wang, and Gregory Abowd

“The AI-Seminar is sponsored by Yahoo!”

February 18th

Seminar: AI / Computational Sustainability Seminar

Speaker: Neo Martinez, Director and President of the Pacific Ecoinformatics and Computational Ecology Lab

Host: Carla Gomes

Bio: Neo Martinez founded and currently directs the Pacific Ecoinformatics and Computational Ecology Lab in Berkeley, California.  He is a computational ecologist that studies the structure and function of whole ecosystems by developing and empirically testing basic and applied theory about complex networks in general and ecological networks in specific.  Director Martinez is an internationally recognized leader in network, computational, ecological, and interdisciplinary science.  His lab’s research integrates biology, math, computer science, economics, and anthropology and his lab’s network visualizations have become widely recognized icons of ecological complexity.  Director Martinez’ many papers published Science, Nature, PNAS, and PLoS Biology among other top journals and books established widely accepted theory explaining the complexity, robustness, and nonlinear dynamics of large ecosystems including a stunning range of biological diversity.  The theory has been corroborated by data from terrestrial and aquatic habitats from throughout the world including paleo-ecosystems over a half billion years old.  The theory has been applied to ecosystems experiencing biodiversity loss and invasion as well as both subsistence and economic exploitation by humans.
Neo received his B.S. in Biology from Cornell, his M.S. in Limnology and Oceanography from the University of Wisconsin at Madison, and his interdisciplinary M.S. and Ph.D. in Energy and Resources from the University of California at Berkeley where he is an Affiliated Faculty and taught graduate coursework on ecological economics. 

Title :  Sustaining Ecological Networks and their Services

Abstract : :  Ecosystems have long been recognized as networks.  Over 150 years ago, Darwin concluded his "Origin of Species" by describing a "tangled bank" of diverse interacting species that "have all been produced by laws acting around us."  Global change such as climate disruption, biodiversity loss, and over exploitation is threatening the ability of these networks to sustain the human health and welfare critically dependent on their continued functioning.  Research on the structure and dynamics of ecological networks have uncovered law-like regularities underlying their dynamic stability and complexity.  Such insights also allow scientists to predict the consequences of global change and help mitigate the damage this change does to the services provided by ecosystems such as food provision and carbon sequestration. Mitigating ecosystem damage and yielding sustainable ecosystem services requires the integration of natural and social sciences such as ecology and economics.  My talk will describe how network and computational sciences have been applied to these and other fundamental interdisciplinary challenges.  The talk will conclude by briefly sketching out a promising future of this research agenda.

“The AI-Seminar is sponsored by Yahoo!”

February 25th

Speaker: Christoph Kirsch, University of Salzburg, Austria (AI-Systems-Robotics Colloquium

Place: **Upson Hall 1st Floor Lounge**

Time: **11:45am - 1:15pm**

Host: Ken Birman/ Ashutosh Saxena

Title :  The Next Frontier of Cloud Computing is in the Clouds, Literally.

Abstract :   Imagine a fleet of autonomously flying high-performance quadrotor helicopters equipped with cameras and laser range finders gathering data for information acquisition tasks such as search-and-rescue missions and environmental monitoring. Inspired by data center cloud computing, the helicopters do not directly execute any mission code but instead work as servers hosting virtual abstractions of networked autonomous vehicles that perform the actual missions. Similar to virtual machines, virtual vehicles are spatially isolated in memory but may also be temporally isolated for real-time performance and even power-isolated for cost accounting. Virtual vehicles can be created and deployed dynamically at flight time and then migrate from one real vehicle to another in order to aggregate information as efficient and fast as possible. The talk begins with a brief overview of our custom-designed quadrotor, which we developed and built entirely from scratch for best performance. We then discuss the potential capabilities and design challenges of software abstractions and systems infrastructure for cloud computing in the clouds. In particular, we discuss the problem and preliminary solutions of providing spatial, temporal, and power isolation of virtual vehicles (or machines) simultaneously.

Joint work with Silviu Craciunas, Andreas Haas, Hannes Payer, Harald Roeck, Andreas Rottmann, Ana Sokolova, Rainer Trummer (University of
Salzburg) and Joshua Love, Raja Sengupta (UC Berkeley)

“The AI-Seminar is sponsored by Yahoo!”

March 4th

Speaker: Matthew Ginsberg, CEO of On Time Systems and partner Green Driver

Host: Bart Selman

Bio: Matthew L. Ginsberg received his doctorate in mathematics from Ox ford in 1980 at the age of 24. He remained on the faculty in Oxford until 1983, doing research in mathematical physics and computer science; during
this period, he wrote a program that was used successfully to trade stock and stock options on Wall Street.
Ginsberg's continuing interest in arti cial intelligence brought him to Stanford in late 1983, where he remained for nine years. In 1992, he founded cirl, the computational intelligence research laboratory at the University of
Oregon. In 1998, he went on to co-found On Time Systems, cirl's commercial spino working in the areas of route and schedule optimization. He is currently the CEO of both On Time Systems and its partner Green Driver.
Ginsberg's present research interests include optimization and satis ability. He is the author of gib, the world's strongest computer bridge player, and numerous publications in these and other areas. He is also the editor of
Readings in Nonmonotonic Reasoning and the author of Essentials of Artificial Intelligence, both published by Morgan Kaufmann.

Title : "Green Driver: AI in a Microcosm"

Abstract :  The Green Driver app is a dynamic routing application for GPS-enabled smartphones. Green Driver combines client GPS data with real-time traffic light information provided by cities to determine optimal routes in response to driver route requests. Routes are optimized with respect to travel time, with the intention of saving the driver both time and fuel, and rerouting can occur if warranted. During a routing session, client phones communicate with a centralized server that both collects GPS data and processes route requests. All relevant data are anonymized and saved to databases for analysis; statistics are calculated from the aggregate data and fed back to the routing engine to improve future routing. Analyses can also be performed to discern driver trends: where do drivers tend to go, how long do they stay, when and where does traffic congestion occur, and so on. The system uses a number of techniques from the field of artificial intelligence.  We apply a variant of A* search for solving the stochastic shortest path problem in order to find optimal driving routes through a network of roads given light-status information. We also use dynamic programming and hidden Markov models to determine the progress of a driver through a network of roads from GPS data and light-status data. The Green Driver system is currently deployed for testing in Eugene, Oregon, and is scheduled for large-scale deployment in Portland, Oregon, in Spring 2011.

"The AI-Seminar is sponsored by Yahoo!"

March 11th

Speaker: Congcong Li (PhD student)

Host:Ashutosh Saxena

Title : Towards Holistic Scene Understanding: Feedback Enabled Cascaded Classification Models

Abstract : Towards Holistic Scene Understanding: Feedback Enabled Cascaded Classification Models

In many domains of artificial intelligence (such as in robotic scene understanding),  several related sub-tasks (such as scene categorization, depth estimation, object detection) operate on the same raw data and provide correlated outputs. Each of these tasks is often notoriously hard, and state-of-the-art classifiers already exist for many sub-tasks.  It is desirable to have an algorithm that can capture such correlation without requiring to make any changes to the inner workings of any classifier.  We propose Feedback Enabled Cascaded Classification Models (FE-CCM),  that maximizes the joint likelihood of the sub-tasks, while requiring only a `black-box'
interface to the original classifier for each sub-task. In this talk we will describe the structure of the FE-CCM and the corresponding learning and inference algorithms. We will also discuss how the model automatically finds connections between tasks and how it handles practical issues such as heterogeneous datasets. We will apply our algorithm to several applications such as scene understanding in computer vision, object grasping in robotics, and so on.

“The AI-Seminar is sponsored by Yahoo!”

March 18th

Speaker: Andrew Tomkins

Host: Jon Kleinberg

Title :  User Modeling on the World Wide Web

Abstract : Online web pages today offer different levels of personalization, from none at all, to simple personalization based on profiles or user-specific data, to complex model-based personalization.  In this talk I discuss the nature and prevalence of pageviews from these different classes.  I also survey improvements due to personalization in search, advertising, recommendations, and content optimization, and describe a series of open research questions in the area.

“The AI-Seminar is sponsored by Yahoo!”

March 25th

Speaker: NO SEMINAR- Spring Break


Bio :



“The AI-Seminar is sponsored by Yahoo!”

April 1st

Speaker: Naren Ramakrishnan. Professor and Associate Head for Graduate Studies

Host: Carla Gomes

*** 2:30pm *** (please note change in time)

Bio: Naren Ramakrishnan is a professor and the associate head for graduate studies in the Department of Computer Science at Virginia Tech. He is also an adjunct professor at the Institute for Bioinformatics and Applied Biotechnology (IBAB) in Bangalore, India. His research interests span data mining in multiple scientific and engineering domains. His work has been featured in the NIH outreach publication Biomedical Computation Review, the National Science Foundation's Discoveries series, and ACM TechNews. Ramakrishnan serves on the editorial boards of many journals including IEEE Computer and Data Mining and Knowledge Discovery.

Title :  Data Mining Techniques for Chiller Management and Lifecycle Assessment

Abstract : How do we cool a data center? Is an e-book reader more environmentally friendly than a paper book? We present two joint projects between Virginia Tech and HP Labs that use data mining techniques to answer questions such as the above. In the area of data centers, we show how temporal data mining techniques can analyze process dynamics in a data center's cooling infrastructure and help identify inefficiencies in operation. In the area of lifecycle assessment, we show how we can reconstruct inventory trees from impact factor databases. This aids in understanding product compositions and in designing environmentally sustainable alternatives. We will also provide some broader perspectives on data mining problems in sustainability.

“This AI / Computational Sustainability seminar is sponsored by the Institute for Computational Sustainability and Yahoo!”

April 8th

Speaker: Eric Xing, CMU

Host: Thorsten Joachims

Bio : Dr. Eric Xing is an associate professor in the School of Computer Science at Carnegie Mellon University. His principal research interests lie in the development of machine learning and statistical methodology; especially for solving problems involving automated learning, reasoning, and decision-making in high-dimensional and dynamic possible worlds; and for building quantitative models and predictive understandings of biological systems. Professor Xing received a Ph.D. in Molecular Biology from Rutgers University, and another Ph.D. in Computer Science from UC Berkeley. His current work involves, 1) foundations of statistical learning, including theory and algorithms for estimating time/space varying-coefficient models, sparse structured input/output models, and nonparametric Bayesian models; 2) computational and statistical analysis of gene regulation, genetic variation, and disease associations; and 3) application of statistical learning in social networks, data mining, vision. Professor Xing has published over 120 peer-reviewed papers, and is an associate editor of the Annals of Applied Statistics, the PLoS Journal of Computational Biology, and an Action Editor of the Machine Learning journal. He is a recipient of the NSF Career Award, the Alfred P. Sloan Research Fellowship in Computer Science, and the United States Air Force Young Investigator Award.

Title :  Reverse Engineering Rewiring Social and Biological Networks Underlying Dynamics Processes

Abstract :  Across the sciences, a fundamental setting for representing and interpreting information about entities, organizations of communities, and changes in these over time, is a stochastic network that is topologically rewiring and semantically evolving over time, space, or genealogy. While there is a rich literature in modeling invariant networks, until recently, little has been done toward modeling the dynamic processes underlying rewiring networks, and on recovering such networks when they are not observable.

In this talk, I present our recent work on estimating temporally varying coefficient and varying structure (VCVS) graphical models, as opposed to invariant model as widely considered in current literature, underlying nonstationary time series data, such as social states of interacting individuals or microarray expression proles of gene networks, along with theoretical results on the asymptotic sparsistency of the proposed estimators. I will then introduce a new Bayesian approach for estimating and visualizing the trajectories of latent multi-functionality of nodal states in these evolving networks. I will show some promising empirical results on recovering and analyzing the latent evolving social networks in the US Senate, and the evolving gene network of fruit fly while aging, at a time resolution only limited by sample frequency. In all cases, our methods reveal interesting dynamic patterns in the networks.

“The AI-Seminar is sponsored by Yahoo!”

April 15th

Speaker: Victor Brodsky, Weill Cornell Medical College

Host: Ramin Zabih

Bio : Dr. Brodsky is an Assistant Medical Director of Informatics and an Assistant Professor in the Departments of Pathology and Laboratory Medicine and Public Health at Weill Cornell Medical College. He serves as the Co-Chair of the Anatomic Pathology Work Group at Health Level 7 (HL7) and is consulting for a multi-institutional NIH U24 grant, “VIVO: Enabling National Networking of Scientists.” Dr. Brodsky is a co-author of multiple research papers on pathology informatics and is currently working with Integrating the Healthcare Enterprise (IHE) on the standardization of the pathology report. He is a member of Diagnostic Intelligence and Health Information Technology (DIHIT) committee of the College of American Pathologists (CAP) and is CAP's liaison to HL7.  Dr. Brodsky is also the current New York State's Empire Clinical Research Investigator Program (ECRIP) fellow.

Title :  AI in Medical Pathology: New opportunities for machine vision and language processing

Abstract :  Recent improvements in storage density and cost are making it possible to transform the field of surgical pathology. Making a cancer diagnosis requires the pathologist to view tissue under the microscope, locate the lesion, and categorize it.  New equipment brings the promise of digitizing tissue images and in turn lets us reexamine the potential role AI can play in this medical field.

“The AI-Seminar is sponsored by Yahoo!”

April 22nd



Speaker: Yun Jiang, PhD Student

Host: Ashutosh Saxena

Title :  Learning in Personal Robotic Manipulation

Abstract :  With recent progress in the robotics technology, we will very soon
have personal robots in our homes and offices. For robots to perform
assistive tasks such as loading and unloading dishwashers, cleaning up
cluttered homes and offices, cooking simple kitchen meals and
assembling furniture, they need to be able to perform the following
two basic manipulation tasks: picking up and placing objects.

In this talk, we address these tasks from a machine learning
perspective, which enables us to tackle variations in unstructured
environments and tackle objects not seen by the robot before. In
detail, given an image and a point cloud from recent RGBD cameras, we
extract 2D and 3D features to capture properties of good grasps and
placements, and use supervised learning algorithms to learn a mapping
from the features to a score for the good grasps and placements.  In
particular, we present algorithms for how to grasp novel objects, how
to infer proper placements for novel objects under different contexts,
and how to arrange multiple objects within limited space. We also show
how to address multiple-object placing as a combination of a learning
problem and a variation of assignment problem.

Joint work with Changxi Zheng, Stephen Moseson, Marcus Lim, Prof.
Ashutosh Saxena

“The AI-Seminar is sponsored by Yahoo!”

April 29th

Speaker: Tony Jebara, Columbia

Host: Thorsten Joachims

Bio : Tony Jebara is Associate Professor of Computer Science at Columbia University and co-founder of Sense Networks. He directs the Columbia Machine Learning Laboratory whose research intersects computer science and statistics to develop new frameworks for learning from data with applications in vision, networks, spatio-temporal data, and text. Jebara has published over 75 peer-reviewed papers in conferences and journals including NIPS, ICML, UAI, COLT, JMLR, CVPR, ICCV, and AISTAT. He is the author of the book Machine Learning: Discriminative and Generative and co-inventor on multiple patents in vision, learning and spatio-temporal modeling. In 2004, Jebara was the recipient of the Career award from the National Science Foundation. His work was recognized with a best paper award at the 26th International Conference on Machine Learning, a best student paper award at the 20th International Conference on Machine Learning as well as an outstanding contribution award from the Pattern Recognition Society in 2001. Jebara's research has been featured on television (ABC, BBC, New York One, TechTV, etc.) as well as in the popular press (New York Times, Slash Dot, Wired, Businessweek, IEEE Spectrum, etc.). He obtained his PhD in 2002 from MIT. Esquire magazine named him one of their Best and Brightest of 2008. 

Title :  Graphical Modeling and Machine Learning with Perfect Graphs

Abstract :  Many graphical modeling and learning problems are NP-hard in general. For instance, maximum a posteriori (MAP) estimation, structured prediction, and marginal inference with graphical models containing cycles can require exponential time in the worst case. Similarly, many combinatorics problems on graphs such as maximum clique, maximum independent set and coloring are also NP-hard. However, a fairly large family of graphs known as perfect graphs admits exact solutions in polynomial time. On perfect graphs, these hard problems are efficiently solvable using linear programming, semidefinite programming, or message passing. We discuss how machine learning and graphical modeling can similarly exploit the perfect graph family. We will present alternative algorithms that solve MAP problems that are otherwise poorly handled by max-product message-passing, dynamic programming and the junction-tree algorithm.

“The AI-Seminar is sponsored by Yahoo!”


May 6th

Speaker: Andy Gallagher (Senior Principal Research Scientist from Kodak)

Host: Ashutosh Saxena

Bio : Andrew Gallagher joined Kodak Research Labs in 1996 initially working on image enhancement algorithms that became Kodak Perfect Touch, embedded in many of Kodak's consumer imaging products. Andy then went back to school, first to Rochester Institute of Technology in 2000 for an M.S., then to Carnegie Mellon University for a Ph.D. in 2009, both in electrical and computer engineering. More recently, Andy has been working with application of images of people and machine learning and is a senior principal scientist with Eastman Kodak. Andy enjoys bicycling, puzzles, whittling, games, and family.

Title :  Solving Jigsaw Puzzles by Measuring Puzzle Piece Compatibility

Abstract :  This talk explores compatibility measures for quantifying pairwise jigsaw piece compatibility. In previous work on jigsaw puzzle assembly, smoothness across potential adjacent jigsaw pieces is encouraged by measuring the differences in color values across the boundary. While this strategy is largely effective, it is subject to fail when gradients or edges pass through boundary of the pair of jigsaw pieces. A new compatibility measure, Mahalanobis Gradient Compatibility, (MGC) is proposed that learns the distribution of gradients on each side of the jigsaw piece boundary, and computes a Mahalanobis distance across the boundary based on this distribution. The new compatibility measure is compared with traditional measures to show a significant improvement at finding the correct match to a particular jigsaw puzzle piece. Further, MGC is shown to be comparable with human performance on the same task. Finally, in practice, the effectiveness of MGC allows us to reconstruct jigsaw puzzles with accuracy that is significantly better than reported in previous work. Perfect reconstruction is achieved on jigsaw puzzles with over 1000 pieces with no anchor patches, more than doubling the number of pieces of previous works.

“The AI-Seminar is sponsored by Yahoo!”

May 13th

Speaker: Cristian Danescu-Niculescu-Mizil & Myle Ott

Host: Lillian Lee

Speaker: Myle Ott

Title :  Finding deceptive opinion spam by any stretch of the imagination

Abstract :  Consumers increasingly rate, review and research products online. Consequently, websites containing consumer reviews are becoming targets of opinion spam. While recent work has focused primarily on manually identifiable instances of opinion spam, in this work we study deceptive opinion spam---fictitious opinions that have been deliberately written to sound authentic. Integrating work from psychology and computational linguistics, we develop and compare three approaches to detecting deceptive opinion spam, and ultimately develop a classifier that is nearly 90% accurate on our gold-standard opinion spam dataset. Based on feature analysis of our learned models, we additionally make several theoretical contributions, including revealing a relationship between deceptive opinions and imaginative writing.

This is joint work with Claire Cardie, Yejin Choi and Jeff Hancock.

Speaker: Cristian Danescu-Niculescu-Mizil

Title: Mark my words! Linguistic style coordination in social media (and movie dialogs)

Abstract:  Participants in conversations tend to converge to one another’s communicative behavior: they coordinate in a variety of dimensions including choice of words, syntax, utterance length, pitch and gestures.  Until now, this psycholinguistic hypothesis has been empirically supported exclusively through small-scale or controlled laboratory studies. Here we address this phenomenon in the context of Twitter conversations.  Undoubtedly, this setting is unlike any other in which coordination was observed. Its novelty comes not only from its size, but also from the non real-time nature of conversations, from the 140 character length restriction, from the wide variety of social relation types, and from a design that was initially not geared towards conversation at all.  Given such constraints, it is not clear a priori whether coordination is robust enough to occur in this new environment.  To investigate this, we develop a probabilistic framework that can model linguistic coordination and measure its effects.  We apply it to a large Twitter conversational dataset specifically developed for this task.  This is the first time the hypothesis of linguistic style coordination has been examined (and verified) in a large scale, real world setting.  Moreover, by investigating the concept of stylistic influence, we are revealing a complexity of the phenomenon which was never observed before.

Furthermore, by also studying linguistic coordination in the context of movie dialogs, we provide some insight into the causal mechanism behind this phenomenon, a topic that has generated substantial scrutiny and debate for over forty years.

This talk includes joint work with Susan Dumais, Michael Gamon and Lillian Lee.

“The AI-Seminar is sponsored by Yahoo!”

May 20th

Speaker: Lars Backstrom, Facebook

Host: Jon Kleinberg

Title : People You May Know: Friend Suggestions on Facebook

Abstract : Facebook's friend recommendation system helps people connect with their friends. Our system, called People You May Know, uses a combination of results from sociology and machine learning to make the best suggestions possible. We will look at some of the challenges involved in building a system that can handle the scale of Facebook and provide high quality recommendations. In this talk I will discuss both the scaling and machine learning challenges that we have faced and overcome in building this system.

“The AI-Seminar is sponsored by Yahoo!”



See also the AI graduate study brochure.

Please contact any of the faculty below if you'd like to give a talk this semester. We especially encourage graduate students to sign up! 

Sponsored by

CS7790, Fall '09
Claire Cardie
Carla Gomes
Joe Halpern
Dan Huttenlocher
Thorsten Joachims
Lillian Lee
Ashutosh Saxena
Bart Selman
Ramin Zabih

Back to CS course websites