We have been conducting behavioral experiments in which human subjects attempt to solve challenging graph-theoretic optimization problems through only local interactions and incentives. The primary goal is to shed light on the relationships between network structure and the behavioral and computational difficulty of different problem types.
To date, we have conducted experiments in which subjects are incentivized to solve problems of graph coloring, consensus, independent set, and an exchange economy game. I will report on thought-provoking findings at both the collective and individual behavioral levels, and contrast them with theories from theoretical computer science, sociology, and economics.
This talk discusses joint work with Stephen Judd, Sid Suri, and Nick Montfort.
Bio: Since 2002, Michael Kearns has been a Professor in the Computer and Information Science Department at the University of Pennsylvania, where he holds the National Center Chair in Resource Management and Technology. He has a secondary appointment in the Operations and Information Management (OPIM) department of the Wharton School, and until July 2006 was the co-director of Penn's interdisciplinary Institute for Research in Cognitive Science. He also leads the Systematic Trading group of Bank of America Securities in New York City. From 1991 until 2001, his research focus was in basic AI and machine learning at AT&T Labs and Bell Labs. During his last four years there, he was the head of the AI department, which conducted a broad range of systems and foundational AI work. Prof. Kearns holds an undergraduate degree from the University of California at Berkeley in Math and Computer Science and a Ph.D. in Computer Science from Harvard University.
Date: 02/18/08
Time: 10am
Location: CoRE Auditorium
Much statistical inference is concerned with controlling some form of tradeoff between flexibility and variability. In Bayesian inference, such control is often exerted via hierarchies---stochastic relationships among prior distributions. Nonparametric Bayesian statisticians work with priors that are general stochastic processes (e.g., distributions on spaces of continuous functions, spaces of monotone functions, or general measure spaces). Thus flexibility is emphasized and it is of particular importance to exert hierarchical control. In this talk I discuss Bayesian hierarchical modeling in the setting of two particularly interesting stochastic processes: the Dirichlet process and the beta process. These processes are discrete with probability one, and have interesting relationships to various random combinatorial objects. They yield models with open-ended numbers of "clusters" and models with open-ended numbers of "features," respectively. I discuss Bayesian modeling based on hierarchical Dirichlet process priors and hierarchical beta process priors, and present applications of these models to problems in biology (statistical genetics, protein structural modeling), information retrieval and computational vision.
Joint work with Yee Whye Teh and Romain Thibaux.
Bio: Michael Jordan is a Professor in the Department of Electrical Engineering and Computer Science and the Department of Statistics at the University of California, Berkeley. He received his Masters from Arizona State University, and earned his PhD in 1985 from the University of California, San Diego. He was a professor at the Massachusetts Institute of Technology from 1988 to 1998. He has published over 250 research articles on topics in computer science, statistics, electrical engineering, molecular biology and cognitive neuroscience. His research in recent years has focused on probabilistic graphical models, kernel machines, nonparametric Bayesian methods and applications to problems in information retrieval, signal processing and bioinformatics. Prof. Jordan was named a Fellow of the American Association for the Advancement of Science (AAAS) in 2006. He is a Fellow of the IMS, a Fellow of the IEEE and a Fellow of the AAAI.
Date: 03/27/08
Time: 10am
Location: CoRE Auditorium
Intel Research's bold "Essential Computing" vision is focused on a leaping capability which will make our interactions with technology intuitive, helpful and robust. These next generation computing systems will move from task and utility-orientation to supporting the essence of our lives. A major challenge is to eliminate annoying, labor-intensive and fragile interactions. Is Inference the key ingredient to make this leap?
We are launching a major focused research effort exploring Inference, combined with power-efficient parallel computing, large scale data, sensors, and wireless communications to make technology more intuitive and robust. These activities are organized into six research themes: Personal Awareness, Richly Communicative, Physicality, Concealing Complexity, Data Rich and Biosensors.
Bio: As a Vice President and Director of Intel Research, Dr. Andrew Chien oversees Intel's exploratory research in seven sites, including Intel's innovative network of university labs, and leadership of Intel's research programs with universities and governments around the world.
Prior to joining Intel, Dr. Chien was the SAIC Endowed Chair Professor in Computer Science and Engineering at the Univ. of California, San Diego and a Senior Fellow at the San Diego Supercomputing Center. He was the founding director of the Center for Networked Systems, a university-industry alliance focused on developing technologies for robust, secure, and open networked systems. For over 20 years, Dr. Chien has led research and development of high-performance computing systems, with expertise in networking, grids, high performance clusters, distributed systems and computer architecture. Dr. Chien is an NSF Young Investigator, ACM Fellow, and IEEE Fellow. Chien serves on the Computing Research Association board of directors, and on the National Science Foundation CISE Director's Advisory Committee, and the Government University Industry Roundtable (GUIRR).
Date: 04/30/08
Time: 10am
Location: CoRE Auditorium