Select a calendar:
Filter November Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Conferences, Lectures, & Seminars
Events for November
-
CS Colloquium
Tue, Nov 09, 2010 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Shaddin Dughmi, Stanford University
Talk Title: How to Compute in a Selfish Society: Randomness May be the Key
Abstract: Algorithmic Mechanism Design is concerned with solving computational problems in situations where essential problem data is being held privately by selfish agents. Techniques from economics have long existed for aligning the incentives of the agents with the social good, yet they often require solving a hard optimization problem exactly. On the other hand, computer scientists have coped with intractability by designing approximation algorithms. Unfortunately, recent results have demonstrated that these two approaches are fundamentally at odds for deterministic mechanisms: combining truthfulness and polynomial-time computation results in an inevitable deterioration of the approximation ratio for many important problems.
Fortunately, there is hope: randomized mechanisms are emerging that reconcile computational and economic constraints, yielding optimal approximate mechanisms for problems where deterministic mechanisms provably fail. In this talk, I will advocate randomized mechanism design by taking a tour through a sequence of our recent results. I will illustrate the power of randomized mechanisms by: (1) Overviewing recent positive results for paradigmatic problems such as multi-unit auctions and variants of combinatorial auctions, and (2) Showing how a black-box reduction can transforms any FPTAS for a social-welfare maximization problem into a truthful FPTAS , and (3) Arguing that, in the future, there is hope for more powerful black box reductions that would yield sweeping positive results for welfare-maximization problems in general.
Biography: Shaddin Dughmi is a PhD student in the computer science theory group at Stanford University, advised by Professor Tim Roughgarden. His interests include algorithms, game theory, and combinatorial optimization. Recently, Shaddin has focused on the following meta-question in algorithmic mechanism
design: When and how can we efficiently compute a desirable solution to a resource allocation problem despite the presence of selfish behavior? Shaddin graduated from Cornell University in 2004 with a B.S. in computer science and a minor in applied mathematics. From 2004 to 2006, he was an Information Security Engineer at the MITRE Corporation, where he worked on cryptographic protocol analysis. He enrolled in the Stanford computer science PhD program in the Fall of 2006, with an expected graduation date of June 2011.
Host: Dr. David Kempe
Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: Kanak Agrawal
-
CS Colloquium
Mon, Nov 15, 2010 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Katerina Argyraki, EPFL
Talk Title: Verifiable Network-Performance Measurements
Abstract: In the current Internet, there is no clean way to troubleshoot poor forwarding performance: when an Internet service provider (ISP) does not forward traffic as agreed/expected, its customers and peers resort to ad-hoc probing to localize and assess the problem. Research proposals advocate end-to-end measurements from different vantage points (e.g., PlanetLab nodes) as a way to forcefully extract information on an ISP's performance without any involvement from the ISP itself. I will argue that it is time to consider a different approach, where ISPs willingly contribute information on their performance, albeit in a way that forces them to tell the truth.
I will present Network Confessional, a system and protocol that enables ISPs to disclose accurate information on their forwarding performance. This information is verifiable -- ISPs cannot manipulate it to significantly exaggerate their performance -- and independently tunable -- each ISP is free to choose its own trade-off between the accuracy of its performance estimates and the resources it devotes to this purpose.
Network Confessional requires deploying modest functionality at participating domains' border routers; I will show that required resources are well within the capabilities of modern networks and can be implemented using today's hardware.
Biography: Katerina Argyraki is a network researcher at EPFL, Switzerland, where she works on programmable networks and techniques for network troubleshooting. She received her Ph.D. in Electrical Engineering from Stanford University in 2007. Her graduate-student years were divided between Stanford's Distributed Systems Group, where she worked on defenses against denial-of-service attacks, and various startups -- Kealia (now part of Sun), BlueArc, and Arista Networks.
Host: Profs. Konstantinos Psounis and Ramesh Govindan
Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
Audiences: Everyone Is Invited
Contact: Kanak Agrawal
-
GTHB Seminar
Tue, Nov 16, 2010 @ 12:00 PM - 02:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Tim Roughgarden, Stanford University
Talk Title: Intrinsic Robustness of the Price of Anarchy
Abstract: The price of anarchy is a measure of the inefficiency of selfish behavior that has been successfully analyzed in many applications, including network routing, resource allocation, network formation, and even models of basketball. It is defined as the worst-case ratio between the welfare of a Nash equilibrium and that of an optimal (first-best) solution. Seemingly, a bound on the price of anarchy is meaningful only if players successfully reach some Nash equilibrium. Our main result is that for many of the classes of games in which the price of anarchy has been studied, results are "intrinsically robust" in the following sense: a bound on the worst-case price of anarchy for pure Nash equilibria *necessarily* implies the exact same worst-case bound for a much larger sets of outcomes, including mixed Nash equilibria, correlated equilibria, and sequences of outcomes generated by natural experimentation strategies (such as successive best responses or simultaneous regret-minimization).
Biography: Tim Roughgarden received his PhD from Cornell University in 2002 and joined the Stanford CS faculty in 2004. His research interests lie in theoretical computer science, especially its interfaces with game theory and networks. He wrote the book "Selfish Routing and the Price of Anarchy" (MIT Press, 2005) and co-edited the book "Algorithmic Game Theory", with Nisan, Tardos, and Vazirani (Cambridge, 2007). His significant awards include the 2002 ACM Doctoral Dissertation Award (Honorable Mention), the 2003 Tucker Prize, the 2003 INFORMS Optimization Prize for Young Researchers, speaking at the 2006 International Congress of Mathematicians, a 2007 PECASE Award, the 2008 Shapley Lectureship of the Game Theory Society, and the 2009 ACM Grace Murray Hopper Award.
Host: GTHB
Location: James H. Zumberge Hall Of Science (ZHS) - 159
Audiences: Everyone Is Invited
Contact: Kanak Agrawal
-
College Commons Panel Discussion
Tue, Nov 16, 2010 @ 04:00 PM - 06:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Sharon Swartz (Evolutionary Biology, Brown University), Akira Lippit (Cinema USC), and Michael Arbib (Computer Science and Neuroscience, USC) , Brown University & USC
Talk Title: Thinking With/As Animals
Abstract: When bees dance, when birds and whales sing and when bats echolocate, how close do these communicative methods come to what we call âlanguageâ? Furthermore, within evolutionary processes, how do manual gestures among humans become speech and how does a leg, in the case of the bat, become a wing? What essential changes to the nature of the human or the animal are signified by speech and flight? And how do we represent the relations between humans and animals in terms of choreographies of the gaze? Why and when do animals look at humans? What do they see when they do look? And how are human and animal gazes the same or different?
In a wide-ranging and dynamic panel discussion between Sharon Swartz (Evolutionary Biology, Brown University), Akira Lippit (Cinema USC), and Michael Arbib (Computer Science and Neuroscience, USC) we will engage these questions and more about the differences and similarities between animals and humans.
To secure your spot please RSVP to: tcc@college.usc.edu
Part IV of a Series of V: THE HUMAN-ANIMAL DIVIDE
Host: College Commons
Location: Edward L. Doheny Jr. Memorial Library (DML) - 240
Audiences: Everyone Is Invited
Contact: Kanak Agrawal
-
CENG, CS & CED/WIE Panel Discussion
Thu, Nov 18, 2010 @ 12:00 PM - 01:30 PM
Thomas Lord Department of Computer Science, Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Mondira (Mandy) Pant, Intel, and Dr. Charles Lee Isbell, Jr., Georgia Tech
Talk Title: Why Pursue Graduate School?
Abstract: This panel encourages students to pursue graduate degree(s) in computing and engineering fields at Masterâs and Ph.D. levels. It aims to inspire and prepare students to be successful in graduate school pursuits. Questions addressed by the panel include the following: Why attend grad school, and why in a computing/engineering field as opposed to some other professional field? How does a graduate degree in a computing/engineering field impact oneâs career opportunities and earning potential? 3) What is the difference between a Masters and PhD, how long do each take, and how do the possible career paths differ between the two degrees? What is exciting about doing research, and how can one find out if research is interesting to him/her? How does one get accepted into graduate school, which schools, and how to pay for it? How can one best prepare him/herself to succeed in grad school? What are the biggest challenges?
Host: Prof. Timothy Pinkston, Senior Associate Dean of Engineering
Location: Ronald Tutor Hall of Engineering (RTH) - 324
Audiences: Everyone Is Invited
Contact: Estela Lopez
-
CS Colloquium: CRA-W/CDC Distinguished Lecture Series
Thu, Nov 18, 2010 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Dr. Charles Lee Isbell, Jr., Georgia Tech
Talk Title: Adaptive Drama Management: Bringing Machine Learning to Interactive Entertainment
Abstract: In recent years, there has been a growing interest in constructing rich interactive entertainment and training experiences. As these experiences have grown in complexity, there has been a corresponding growing need for the development of robust technologies to shape and modify those experiences in reaction to the actions of human participants.
When thinking about how machine learning and artificial intelligence could help, one notes that the traditional goal of AI games---to win the game---is not particularly useful; rather, the goal is to make the human player's play experience better while being consistent with the goals of the author.
In this talk, I will present our technical efforts to achieve this goal by using machine learning as a way to allow designers to specify problems in broad strokes while allowing a machine do further fine-tuning. In particular, I discuss (1) Targeted Trajectory Distribution Markov Decision Processes (TTD-MDPs), an extension of MDPs that provide variety of experience during repeated execution and (2) computational influence, an automated way of operationalizing theories of influence and persuasion from social psychology to help guide players without decreasing their feelings of autonomy. I also describe our evaluation of these techniques with both simulations and an interactive storytelling system with human subjects.
Biography: Dr. Charles Lee Isbell, Jr., received his BS in computer science in 1990 from the Georgia Institute of Technology and his PhD in 1998 from the Massachusetts Institute of Technology. After four years at AT&T Labs, he returned to Georgia Tech as faculty at the College of Computing. Charles' research interests are varied, but recently he has been building autonomous agents that engage in life-long learning in the presence of thousands of other intelligent agents, including humans. His work has been featured in the popular media, including The New York Times and the Washington Post, as well as in technical collections, where he has won two best paper awards in this area. Charles also pursues reform in CS education. He was a developer of Threads, Georgia Tech's new structuring principle for computing curricula. Recently, he has become the Associate Dean of Academic Affairs for the College of Computing.
Host: Dr. Timothy Pinkston, Senior Associate Dean of Engineering
Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: Kanak Agrawal
-
CS Colloquium
Fri, Nov 19, 2010 @ 03:00 PM - 04:30 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Prof. Victor Lesser, University of Massachusetts, Amherst
Talk Title: Reflections on being an AI System Architect
Abstract: I will share with you the intellectual intuitions and serendipities that have shaped my research career. I first discuss my early research that includes my PhD thesis work at Stanford on a reconfigurable multiprocessor and my post-doc work as the system architect for the Hearsay-II system at CMU (the first fully instantiated blackboard system) that have strongly influenced my later research. These ideas will include distribution of control, meta-level and self-aware control, managing inconsistency rather than eliminating it, the importance of learning as an integral part of a system's architecture, and recognizing that experimentation is more than gathering statistics. In discussing these ideas, I will present a number of systems that I have developed with my students that embody these ideas. I will conclude the lecture by discussing some of my recent work on organizational control that brings many of these ideas together. The basis of this lecture comes out of two papers on my web site: ftp://mas.cs.umass.edu/pub/lesser/system_architect_webdoc.pdf and ftp://mas.cs.umass.edu/pub/LabHistory_Web-Article.pdf
Biography: Victor R. Lesser received his B.A. in Mathematics from Cornell University in 1966, and the Ph.D. degree in Computer Science from Stanford University in 1973. He then was a post-doc/research scientist at Carnegie-Mellon University, working on the Hearsay-II speech understanding system. He has been a professor in the Department of Computer Science at the University of Massachusetts Amherst since 1977, and was named Distinguished Professor of Computer Science in 2009. His major research focus is on the control and organization of complex AI systems. He is considered a leading researcher in the areas of blackboard systems, multi-agent/ distributed AI, and real-time AI. He has also made contributions in the areas of computer architecture, signal understanding, diagnostics, plan recognition, and computer-supported cooperative work. He has worked in application areas such as sensor networks for vehicle tracking and weather monitoring, speech and sound understanding, information gathering on the internet, peer-to-peer information retrieval, intelligent user interfaces, distributed task allocation and scheduling, and virtual agent enterprises.
Professor Lesser is a Founding Fellow of the Association for the Advancement of Artificial Intelligence (AAAI) and an IEEE Fellow. He was General Chair of the first international conference on Multi-Agent Systems (ICMAS) in 1995, and Founding President of the International Foundation of Autonomous Agents and Multi-Agent Systems (IFAAMAS) in 1998. To honor his contributions to the field of multi-agent systems, IFAAMAS established the "Victor Lesser Distinguished Dissertation Award." He received the UMass Amherst College of Natural Sciences and Mathematics (NSM) Outstanding Teaching Award (2004) and Outstanding Research Award (2008), and the Chancellor's Award for Outstanding Accomplishments in Research and Creative Activity (2008). Professor Lesser was also the recipient of the IJCAI-09 Award for Research Excellence.
Host: Prof. Milind Tambe
Location: Seeley G. Mudd Building (SGM) - 124
Audiences: Everyone Is Invited
Contact: Kanak Agrawal
-
CS Colloquium
Tue, Nov 23, 2010 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Dr. Chun-Nan Hsu, Information Sciences Institute (ISI)
Talk Title: Accelerating Machine Learning by Aggressive Extrapolation
Abstract: This talk presents how to accelerate statistical machine learning algorithms for large scale applications by aggressive extrapolation. Extrapolation methods, such as Aitken's acceleration, have the advantage that they can achieve quadratic convergence with an overhead linear to the dimension of the training data. However, they can be numerically unstable and their convergence is only locally guaranteed. We show that this can be fixed by a double extrapolation method. There are two options for the extrapolation, global or component-wise. Previously, it was not clear which option is more effective. We show a general condition to determine which option will be more effective and show how to apply the condition to the training of Bayesian networks and conditional random fields (CRF). Then we show that extrapolation can accelerate on-line learning with a method called Periodic Step-size Adaptation (PSA). We show that PSA is an approximation of a theoretic "single-pass" on-line learning method, which can converge to an empirical optimum in a single pass through the training examples. With a single-pass on-line learning method, disk I/O can be minimized when a training set is too large to fit in memory. Experimental results for a wide variety of models, including CRF, linear SVM, and convolutional neural networks, show that single-pass performance of PSA is always very close to empirical optimum. Finally, an application to gene mention tagging for biological text mining will be presented, which achieved the top score in BioCreative 2 challenge in 2007 and again in BioCreative 3 challenge in 2010.
Biography: Dr. Chun-Nan Hsu is a computer scientist at Information Sciences Institute (ISI). Prior to joining ISI, he is Research Fellow and Leader of the Adaptive Internet Intelligent Agents (AIIA) Lab at the Institute of Information Science, Academia Sinica, Taipei, Taiwan. His research interests include machine learning, data mining, databases and bioinformatics. He earned his M.S. and Ph.D. degree in Computer Science from the University of Southern California, Los Angeles, CA, in 1992 and 1996, respectively. In 1996, before he passed his doctoral oral exam, he had been offered a position as Assistant Professor at the Department of Computer Science and Engineering, Arizona State University, Tempe, AZ. He taught there for two years before he returned to Taiwan in 1998. Since 2005, he has been the principal investigator of the Advanced Bioinformatics Core, National Research Program in Genomic Medicine, Taiwan, and leading one of the largest research efforts in computerized drug design and discovery in Taiwan. In 2006, the first drug candidate due to the use of the software his team developed was commercialized. In 2007, his teams achieved the best scores in the BioCreative 2 text mining challenge. Dr. Hsu has published about 90 scientific articles since 1993. Currently, Dr. Hsu has been working on applying artificial intelligence to computational biology and bioinformatics.
Host: Dr. Dennis McLeod
Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: Kanak Agrawal
-
CS Colloquium
Tue, Nov 23, 2010 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Dr. Chun-Nan Hsu, ISI, USC, Machine Learning, Information Integration and Bioinformatics
Talk Title: Accelerating Machine Learning by Aggressive Extrapolation
Abstract: This talk presents how to accelerate statistical machine learning algorithms for large scale applications by aggressive extrapolation. Extrapolation methods, such as Aitken's acceleration, have the advantage that they can achieve quadratic convergence with an overhead linear to the dimension of the training data. However, they can be numerically unstable and their convergence is only locally guaranteed. We show that this can be fixed by a double extrapolation method. There are two options for the extrapolation, global or component-wise. Previously, it was not clear which option is more effective. We show a general condition to determine which option will be more effective and show how to apply the condition to the training of Bayesian networks and conditional random fields (CRF). Then we show that extrapolation can accelerate on-line learning with a method called Periodic Step-size Adaptation (PSA). We show that PSA is an approximation of a theoretic "single-pass" on-line learning method, which can converge to an empirical optimum in a single pass through the training examples. With a single-pass on-line learning method, disk I/O can be minimized when a training set is too large to fit in memory. Experimental results for a wide variety of models, including CRF, linear SVM, and convolutional neural networks, show that single-pass performance of PSA is always very close to empirical optimum. Finally, an application to gene mention tagging for biological text mining will be presented, which achieved the top score in BioCreative 2 challenge in 2007 and again in BioCreative 3 challenge in 2010.
Biography: Dr. Chun-Nan Hsu is a computer scientist at Information Sciences Institute (ISI). Prior to joining ISI, he is Research Fellow and Leader of the Adaptive Internet Intelligent Agents (AIIA) Lab at the Institute of Information Science, Academia Sinica, Taipei, Taiwan. His research interests include machine learning, data mining, databases and bioinformatics. He earned his M.S. and Ph.D. degree in Computer Science from the University of Southern California, Los Angeles, CA, in 1992 and 1996, respectively. In 1996, before he passed his doctoral oral exam, he had been offered a position as Assistant Professor at the Department of Computer Science and Engineering, Arizona State University, Tempe, AZ. He taught there for two years before he returned to Taiwan in 1998. Since 2005, he has been the principal investigator of the Advanced Bioinformatics Core, National Research Program in Genomic Medicine, Taiwan, and leading one of the largest research efforts in computerized drug design and discovery in Taiwan. In 2006, the first drug candidate due to the use of the software his team developed was commercialized. In 2007, his teams achieved the best scores in the BioCreative 2 text mining challenge. Dr. Hsu has published about 90 scientific articles since 1993. Currently, Dr. Hsu has been working on applying artificial intelligence to computational biology and bioinformatics.
Host: Prof. Dennis McLeod
Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: Kanak Agrawal
-
CS Colloquium
Tue, Nov 30, 2010 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Dr. Morteza Dehghani , ICT, USC
Talk Title: Investigating and Modeling the Role of Cultural Narratives in Moral Decision-Making
Abstract: In dealing with conflict, two broadly different approaches to modeling the values that drive decisions and choice of behavior have emerged: a consequentialist approach based on instrumental or material values, versus a deontological approach based on moral or sacred values. Sacred values are different from secular values in that they are often associated with violations of the cost-benefit logic of rational choice models. Understanding and modeling the impacts of sacred values on decision making is especially important in resolving intergroup conflicts and negotiations. In this talk, I first examine whether the principles of analogical retrieval and mapping govern the processes by which cultural and sacred narratives are applied. To understand and model this process computationally, I have developed MoralDM as a model of recognition-based moral decision-making. This model relies on a combination of first-principles reasoning and analogical reasoning to model the recognition-based mode of decision making. To discuss the broader impact of the role of narratives on decision making, I examine Iran's stance on its national nuclear program, using it as an indicator of how sacred values can emerge from sacred rhetoric. Overall, I argue that understanding sacred values and the processes by which they emerge are vital for understanding and modeling decision-making in cultural contexts.
Biography: Morteza Dehghani is currently a Research Scientist at Institute of Creative Technologies (ICT) at University of Southern California. Before joining ICT, he was a postdoctoral researcher in the Department of Psychology at Northwestern University and a Young Investigator fellow at ARTIS. His research interests include computational social sciences, cross cultural differences in moral decision making, analogical and case-based reasoning, and cognitive modeling of different aspects of cognition. He is specifically interested in the role of cultural products in decision making and in the emergence of sacred values. His research approach consists of both conducting psychological experiments and computational cognitive modeling. He received his Ph.D. and MS in Computer Science with specialization in Cognitive Science from Northwestern University and MS and BS from University of California at Los Angeles.
Host: Prof. Ewa Deelman
Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: Kanak Agrawal
