Conferences, Lectures, & Seminars
Events for October
-
CS Colloquium (Venue changed)
Wed, Oct 01, 2008 @ 01:30 PM - 02:50 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Relational Agents: Social and Communicative Competencies for Maintaining Engagement with Users Over Multiple InteractionsSpeaker: Prof. Timothy Bickmore, Northeastern UniversityHost: Prof. Maja MataricAbstract:Many applications in healthcare, education, sales and games require maintenance of user adherence to a desired interaction usage pattern. In this talk I will present principles and techniques from a number of disciplines for building and maintaining social bonds between users and computer agents. I will discuss applications of these principles to two very different application areas: health communication by a virtual nurse agent for low health literacy patients, and direction giving by an animated robotic tour guide agent in a science museum.Ninety million Americans have low health literacy, resulting in difficulty reading and following written medical instructions. Evidence suggests that face-to-face encounters with a health provider —in conjunction with written instructions—remains one of the best methods for communicating health information to these individuals. I will describe recent work my lab has done in studying human experts explaining written health instructions to individuals with varying degrees of health literacy, and models of the observed verbal and nonverbal behavior that we have incorporated into computer animated agents that can explain health documents to users. I will present results from a series of lab studies on the efficacy of these agents, in addition to two rounds of user testing of a virtual nurse that performs bedside patient education prior to hospital discharge at Boston Medical Center.I will also discuss the development and evaluation of an animated tour guide agent that has been running in the Boston Museum of Science since April that has interacted with over 20,000 visitors. The main challenges we confronted in developing this agent were sensing visitor conversational cues in a very noisy public environment, and re-identifying repeat visitors (using biometrics) so that prior discourse and relational models could be continued.Biography:Timothy Bickmore is an Assistant Professor in the College of Computer and Information Science at Northeastern University. Dr. Bickmore's research focus is on the development of Relational Agents--computational artifacts designed to build long-term social-emotional relationships with their users. These agents have been deployed within the context of behavior change interventions in which they are designed to establish working alliance relationships with patients in order to maximize intervention outcomes. Prior to joining Northeastern, Dr. Bickmore was an Assistant Professor of Medicine at the Boston University School of Medicine. Dr. Bickmore received his PhD from the MIT Media Lab, studying under Profs. Rosalind Picard (Affective Computing) and Justine Cassell (Gesture and Narrative Language).
Location: Von Kleinsmid Center For International & Public Affairs (VKC) - 150
Audiences: Everyone Is Invited
Contact: CS Colloquia
-
Walking Your Dog in the Woods in Polynomial Time
Thu, Oct 02, 2008 @ 04:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Shripad Thite, CalTech
Host: Prof. David KempeAbstract:
The Frechet distance between two curves in the plane is the minimum length of a leash that allows a dog and its owner to walk along their respective curves, from one end to the other, without backtracking. We propose a natural extension of Frechet distance to more general metric spaces, which requires the leash itself to move continuously over time. For example, for curves in the punctured plane, the leash cannot pass through or jump over obstacles (``trees''). Thus, we introduce the homotopic Frechet distance between two curves embedded in a general metric space. We describe a polynomial-time algorithm to compute the homotopic Frechet distance between two given polygonal curves in the plane minus a given set of polygonal obstacles. Our algorithm produces a walk that minimizes the maximum leash length.This is joint work with Erin Wolf Chambers, Eric Colin de Verdiere, Jeff Erickson, Sylvain Lazard, and Francis Lazarus, which appeared at SoCG'08 in June and was invited to a CGTA special issue.Biography:
Shripad is a postdoctoral fellow in the Center for the Mathematics of Information (CMI) at Caltech. His research is in algorithms, specifically in computational geometry and topology. He designs algorithms for fundamental problems in computational geometry as well as algorithms for geometric problems in applied areas, including scientific computing, graphics and visualization, wireless networking, robotics, and even computational economics. Shripad earned his Ph.D. in Computer Science from the University of Illinois at Urbana-Champaign with Jeff Erickson, and then spent two years at the Technische Universiteit Eindhoven in the Netherlands working with Mark de Berg, before joining Caltech last year.Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: CS Colloquia
-
Multi-Armed Bandits in Metric Spaces
Mon, Oct 06, 2008 @ 11:00 AM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Alex Slivkins, Microsoft Research - SVC
Host: Prof. David KempeAbstract:
In a multi-armed bandit problem, an online algorithm chooses from a set of strategies in a sequence of trials so as to maximize the total payoff of the chosen strategies. While the performance of bandit algorithms with a small finite strategy set is quite well understood, bandit problems with large strategy sets are still a topic of very active investigation, motivated by practical applications such as online auctions and web advertisement. The goal of such research is to identify broad and natural classes of strategy sets and payoff functions which enable the design of efficient solutions.In this work we study a very general setting for the multi-armed bandit problem in which the strategies form a metric space, and the payoff function satisfies a Lipschitz condition with respect to the metric. We refer to this problem as the "Lipschitz MAB problem". We present a complete solution for the multi-armed problem in this setting. That is, for every metric space (L,X) we define an isometry invariant which bounds from below the performance of Lipschitz MAB algorithms for X, and we present an algorithm which comes arbitrarily close to meeting this bound. Furthermore, our technique gives even better results for benign payoff functions. (STOC'08: joint work with Bobby Kleinberg and Eli Upfal)Biography:
Alex Slivkins is a researcher at Microsoft Research, Silicon Valley Center. Before that he has been an undergrad at Caltech, a grad student at Cornell CS dept, and a postdoc at Brown. He is interested in the design and analysis of algorithms, the theory of large distributed networks, and machine learning. Specific topics of interest include metric embeddings, locality-aware overlay networks, and multi-armed bandits.Location: James H. Zumberge Hall Of Science (ZHS) - 163
Audiences: Everyone Is Invited
Contact: CS Colloquia
-
Pitfalls and Paradoxes in the History of Probability Theory
Tue, Oct 21, 2008 @ 04:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Dr. Mike Shlesinger, Office of Naval Research
Host: Prof. Ram NevatiaAbstract:
We trace the history of probability theory from the throwing of bones, sticks, and dice to modern times. Early 18th century books, Jacob Bernouill's "The Art of Conjecture" and Abraham DeMoive's "The Doctrine of Chances" were rich with new mathematics, insight and gambling odds. Progress was often made by confronting paradoxes. The first of these confused probabilities with expectations and was explained in the Pascal-Fermat letters of 1654. The St. Petersburg Paradox involved a distribution with an infinite first moment, and Levy discovered a whole class of probabilities with infinite moments that have found a surprising utility in physics. The Bertrand paradox involved measure theory for continuous probabilities, Poisson discovered that adding random variables need not always produce the Gaussian, and Daniel Bernoulli and D'Alembert argued over the probabilities for the safety of smallpox vaccinations. Using these and other anecdotes, we discuss vignettes that have brought us to our modern understanding of probability theory.Biography:
Dr. Shlesinger holds the Kinnear Chair in Science at the US Naval Academy. At the Office of Naval Research, Dr. Shlesinger is the Research Division Director in the Expeditionary Maneuver Warfare and Combating Terrorism Department. He is also the Program Manager for the Counter-IED basic research program and ONR's Chief Scientist for Nonlinear Science. Dr. Shlesinger joined ONR in 1983 and became a member of the Senior Executive Service in 1987. He is a Fellow of the American Physical Society and has published about 200 scientific papers on topics in stochastic processes, glassy materials, proteins, neurons, and nonlinear dynamics. He is a Divisional Associate Editor of the Physical Review Letters. Dr. Shlesinger received ONR's 2006 Saalfeld Award for Outstanding Lifetime Achievement in Science, the federal government's Presidential Rank Award for Meritorious Senior Professionals in 2004, U. Maryland's Distinguished Postdoc Alum award in 2004, the Navy Superior Civilian Service Award in 1991 and was the Regents' Lecturer at UCSD in 1994, and the Michelson Lecturer at the USNA in 1992. His Ph. D., in Physics, is from the U. of Rochester in 1975, and he holds a B.S. in Mathematics and Physics from SUNY Stony Brook in 1970.Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: CS Colloquia
-
Parsing Images (Distinguished Lecture)
Thu, Oct 30, 2008 @ 04:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Prof. Jitendra Malik, UC Berkeley
Host: Prof. Ram NevatiaAbstract:
When humans look at an image or a video clip, they start from the raw input, a collection of pixels, and infer considerable structure and semantics about the world that is projected into the image. They construct a hierarchical partition of the image into sets of pixels that correspond to "objects" or "parts" of objects, and attach concepts - "dog", "forest" etc to various levels in this hierarchy. We have considerable evidence from perception that this process is based on bottom up cues such as similarity of pixel brightness, color, texture and motion as well as top down input derived from familiar visual categories such as faces or street scenes. Constructing a computational model for this is perhaps the central problem in both human and machine vision. Various subproblems of this grand challenge include image segmentation, perceptual grouping and visual recognition.My research group and I have been studying different aspects of this problem for more than a decade, and at this stage, we feel we have the outlines of a framework for solving it. We start with a local process of marking contours in images based on local differences in brightness, color, texture etc, move on to a more global framework for extracting coherent regions, which in turn help drive a process of visual recognition, which then feedback to refine the grouping itself. We can quantify the performance of the framework on various standard datasets for the subproblems. Of course, much more work needs to be done to attain human level performance, but I feel optimistic that computer vision is on track towards closing the much-cited ``semantic gap'' between pixels and perception.Biography:
Jitendra Malik was born in Mathura, India in 1960. He received the B.Tech degree in Electrical Engineering from Indian Institute of Technology, Kanpur in 1980 and the PhD degree in Computer Science from Stanford University in 1985. In January 1986, he joined the university of California at Berkeley, where he is currently the Arthur J. Chick Professor in the Computer Science Division, Department of Electrical Engg and Computer Sciences. He is also on the faculty of the Cognitive Science and Vision Science groups. During 2002-2004 he served as the Chair of the Computer Science Division and during 2004-2006 as the Department Chair of EECS. He serves on the advisory board of Microsoft Research India, and on the Governing Body of IIIT Bangalore.His current research interests are in computer vision, computational modeling of human vision and analysis of biological images. His work has spanned a range of topics in vision including image segmentation, perceptual grouping, texture, stereopsis and object recognition with applications to image based modeling and rendering in computer graphics, intelligent vehicle highway systems, and biological image analysis. He has authored or co-authored more than a hundred and fifty research papers on these topics, and graduated twenty-five PhD students who occupy prominent places in academia and industry.He received the gold medal for the best graduating student in Electrical Engineering from IIT Kanpur in 1980, a Presidential Young Investigator Award in 1989, and the Rosenbaum fellowship for the Computer Vision Programme at the Newton Institute of Mathematical Sciences, University of Cambridge in 1993. At UC Berkeley, he was selected for the Diane S. McEntyre Award for Excellence in Teaching in 2000, a Miller Research Professorship in 2001, and appointed to be the Arthur J. Chick Professor in 2002. He received the Distinguished Alumnus Award from IIT Kanpur in 2008. He was awarded the Longuet-Higgins Prize for a contribution that has stood the test of time twice, in 2007 and in 2008. He is a fellow of the IEEE.Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: CS Colloquia
