Colloquium Series Fall 2003

Rajat Todi

An execution-driven microarchitecture-accurate microprocessor simulator requires a complex software program. The simulator must be highly detailed and accurate if it is used for microarchitecture design evaluation. The detail and accuracy come at the high cost of enormous simulation time. A simulator that models a modern super-scalar processor is 100,000 to 1,000,000 times slower than the actual hardware being modeled. Running a benchmark in full microarchitecture mode (UA) can be execution time prohibitive. Various methods of sampling the execution have been used to reduce the simulation time. These include: random sampling, and reduced input datasets. On SPEC's CPU95, these sampling methods have successfully reduced the simulation time by factors of 50 to 100 times without a significant reduction in accuracy. With the introduction of SPEC's CPU2000 (SPEC2K) the situation has become more difficult. The datasets and execution times are much longer. Even with sampling, simulation of a single benchmark in the SPEC2K suite takes several days to complete. Moreover, using such methods it is uncertain whether the sampling method is exercising the program regions responsible for bottlenecks in the simulated design. This can lead to inaccurate projected results, thus leading to inaccuracy in design evaluation. Hence, using the above methods, simulators are less effective than they could be due to slowness and inaccuracy in the simulated result.

Joseph P. Bigus

Slides for this talk are provided

The IBM Agent Building and Learning Environment (ABLE) project at IBM T.J. Watson Research Center is a productivity toolkit for building intelligent software applications, including business rules, autonomic computing, and multiagent systems. ABLE provides a lightweight Java agent framework, a comprehensive JavaBeans library of machine learning and reasoning components, a set of Swing and Eclipse development and test tools, and a secure distributed multiagent platform. ABLE is a core IBM autonomic computing technology and has been available for use by the research community via the IBM alphaWorks site since May 2000.

In this talk, we will present an overview of the technology provided in the ABLE toolkit, focussing on two areas: the rules environment and the agent framework and platform. We will describe the ABLE rule language, rule editor, template support, and source level debugger. ABLE features a pluggable rule engine architecture and supports procedural scripting, simple forward and backward chaining, fuzzy systems, a Prolog engine, and a Rete' net pattern matching engine. Furthermore, the rule language can be used as a scripting language to define agent behavior. We will demonstrate the Eclipse agent editor as well as a web-based interface for generation of ABLE rules and rulesets from templates. The ABLE distributed agent platform is based on the FIPA abstract architecture and features an administrative console GUI, conversational support, agent logging, and a secure infrastructure for multiagent applications. Usage of ABLE in applications will be described. Ongoing research and future plans will be discussed.

ABLE technology is shipping with multiple IBM products and is available for use as a testbed for Autonomic and On-Demand computing research and applications. More information can be obtained at the IBM alphaworks site at http://alphaworks.ibm.com/tech/able.

Ramesh Jain

Computing has changed significantly over the last few decades. Computing systems that initially dealt with data and computation rapidly moved to information and communication. The next step on the evolutionary scale is insight and experience. Applications, ranging from education to entertainment and from healthcare to business, are demanding the use of live, spatio-temporal, heterogeneous data in experiential environment. Most computing techniques in use today were developed to meet the information-centric needs of the last quarter of the 20th century. To keep in pace with evolving applications, we need to design experiential environments that let users apply their senses to observe data and information about an event and to interact with aspects of the event that are of particular interest. These environments are essential to explore large volumes of heterogeneous, multifarious, spatio-temporal data to gain insights in the situation through experiencing the data. We discuss requirements of experiential environments and present some approaches to realize such systems.

Joseph Halpern

Please also go to International Symposium on Modern Computing for details.

Uncertainty is a fundamental---and unavoidable---feature of daily life. In order to reason about uncertainty, we need to have the tools to represent it well. I'll discuss a general framework, that incorporates knowledge, time, and probability, and gives us a systematic way of representing uncertainty. In this framework, we can reasoning about agents' beliefs about the world, their beliefs about other agents' beliefs, and their beliefs about the future. Just as importantly, the framework gives us the tools to model at a semantic level many phenomena of interest, from adversarial games to belief revision. Finally, the framework emphasizes the importance of specifying clearly the protocol that generates a system, and provides the tools for describing protocols and the systems they generate. I'll show by example how the framework can give insight into a wide range of problems, from coordination to knowledge base queries to puzzles like the Monty Hall puzzle.

Joseph Halpern

In designing and running systems, decisions must always be made: When should we garbage collect? Which transactions should be aborted? How big should the page table be? How often should we resend a message that is not acknowledged? Currently, these decisions seem to be made based on intuition and experience. There is a whole body of research on decision theory that attempts to provide the basis for making these decisions. I argue that these tools should be taken far more seriously. I illustrate this point by considering a number of examples from distributed computing (the problem of reliable communication) and database systems (join strategies).

James Hendler

Please also go to International Symposium on Modern Computing for details.

The World Wide Web is often referred to as a web of information, but is it? When you ask a query on the web you get pointers to pages, not answers. If you're looking for something beyond text, you're often unable to find it. The next generation of the Web, already in the works, aims to fix this by making more of the content on the web "understandable" to the programs that help us find, filter and use what is out there. In this talk, I will describe this new generation of the web, discuss some of the technologies that will help to power it, and consider some of the ways in which it may be used to create new and powerful web applications beyond the capabilities of the current web. I will also discuss future directions for Semantic Web research.

Pierre Baldi

Please also go to International Symposium on Modern Computing for details.The Schedule is here.

Predicting protein structure is a fundamental problem in biology, especially in the genomic era where over one third of newly discovered genes have unknown structure and function. Because sequence and structure data (hence training sets) continue to grow exponentially, this area is ideally suited for machine learning approaches. Neural networks, in particular, have had remarkable success and have led, for instance, to the construction of the best secondary structure predictors. We will provide an overview of the machine learning state-of-the-art for several structure prediction problems including:

(1) prediction of protein secondary structures; (2) prediction of relative solvent accessibility; (3) prediction of contacts; (4) prediction of three-dimensional protein structures; (5) prediction of interchain beta-sheet quaternary structures; using machine learning methods. The resulting predictors are available at www.igb.uci.edu/tools.htm. While training can take several weeks, once trained the predictors can be used on a proteomic scale. The methods we have developed are based on the theory of graphical models but use deterministic recursive neural networks to speed up learning. We will discuss their applicability to other problems and the lessons learnt for the design of complex neural architectures.

Michael Pazzani

Please also go to International Symposium on Modern Computing for details. The Schedule is here.

Knowledge discovery in databases is a field whose goal is to turn data into knowledge. For example, by analyzing a database of credit card customers we can determine what types of customers are most likely to be profitable for the company. By "mining" databases of medical records, new cost-effective procedures for screening for diseases may be uncovered. We review advances in the field over the past two decades of research in statistics, neural networks and artificial intelligence that have identified a variety of approaches that produce accurate descriptive or predictive models. However, we show that experts are unwilling to accept the results of these techniques when they don't make sense, are difficult to understand, or violate prior understanding. We discuss factors that make learned knowledge acceptable to experts and discuss modifications to rule learning, linear regression and text classification algorithms that make the learned models more comprehensible.

Stephen Smale

Please also go to International Symposium on Modern Computing for details.The Schedule is here.

Sampling theory is a part of mathematics that deals with areas such as signal processing, pattern recognition and even aspects of the internet. we will present new estimates which take into account noise and are associated to classical algorithms of linear algebra.

Thad Starner

Please also go to International Symposium on Modern Computing for details.The Schedule is here.

As computers have evolved in size from mainframes to desktops, their interfaces have become correspondingly more interactive and personal. Currently, another physical change is underway, placing computational power on the user's body. These wearable machines encourage new applications that were formerly infeasible and, correspondingly, will result in new usage patterns.

A fundamental improvement offered by wearable computing is an increased sense of user context. Wearable computers that "see" as the user sees and "hear" as the user hears provide a unique "first-person" viewpoint of the user's environment. By exploiting models of user actions recovered by these systems, intelligent agents and augmented realities can be designed to create almost an intellectual symbiosis between computer and user.

Raghu Ramakrishnan

Please also go to International Symposium on Modern Computing for details.The Schedule is here.

Exploratory data analysis is typically an iterative, multi-step process in which data is cleaned, scaled, integrated, and various algorithms are applied to arrive at interesting insights. Most algorithmic research has concentrated on algorithms for a single step in this process, e.g., algorithms for constructing a predictive model from training data. However, the speed of an individual algorithm is rarely the bottleneck in a data mining project. The limiting factor is usually the difficulty of understanding the data, exploring numerous alternatives, and managing the analysis process and intermediate results. The alternatives include the choice of mining techniques, how they are applied, and to what subsets of data they are applied, leading to a rapid explosion in the number of potential analysis steps.

An important direction for research is how to provide better support for multi- step analyses. I will discuss this problem, and describe a novel framework called subset mining that we are developing and evaluating in the application domain of atmospheric aerosol analysis. As the name suggests, the goal is to identify subsets of data that are 'interesting', or have interesting relationships between them. The use of the approach for multi-step mining is based on the idea that each step in an analysis is focused on some subset of the data, and that multi-step analyses can be understood and composed in terms of the underlying data subsets and their properties.

Satinder Singh

The use of Markov decision process (MDP) models to represent agent-environment interaction has been very fruitful for reinforcement learning and for artificial intelligence in general. After briefly reviewing some of these "fruits" I will discuss the limitations of MDP models and the need to go beyond them. The standard extension of MDPs to partially-observable MDPs, or POMDPs, haven't served us well, at least so far. In this talk, I will present predictive state representations, or PSRs, a new class of predictive models for reinforcement learning. The key idea in PSRs is to use predictions of observable outcomes of tests or experiments the agent can do in its environment to represent the state of the environment. I will show that PSRs are more general than POMDPs and yet are at least as, and often more, compact than POMDPs. I will also present some results on learning PSR models from data and conclude with some reasons for optimism about PSR models as well as with directions for future work on PSRs.

Rich Sutton

Please also go to International Symposium on Modern Computing for details.The Schedule is here.

The primary goal of artificial intelligence, and of cognitive science generally, is to develop a computational theory of mind, that is, of *what* minds compute and *why*, without necessarily saying *how*. Although this view was clearly articulated by David Marr in 1975, and has become standard, there remains precious little in cognitive science that can be seen as computational theory of any generality. Reinforcement learning, an approach to artificial intelligence now about 20 years old, and which has been applied in such diverse areas as process control, robotics, game-playing, operations research, animal learning, and neural reward systems, can be viewed as the beginnings of a computational theory of mind. In this talk, we present the central ideas of this computational theory, which are:

1) The reward hypothesis - that a mind's goal is the maximization of a received scalar signal (reward) 2) The value-function hypothesis - that this goal is achieved by estimating a mapping from states to the total expected reward following them (the value function) And, more speculatively, 3) The prediction hypothesis - that these goals are achieved by learning and composing predictions of "what leads to what." These predictions can be thought of as knowlegde and their composition as reasoning 4) The predictive-state hypothesis - that the state of the world is itself a set of predictions

Prem Uppuluri

Solutions to many important operating system problems are based on extending/enhancing operating system functionality. However, this is a difficult and expensive task as modern operating systems are very large and complex -- running into millions of lines of code which have been developed over many years by a large number of programmers. An alternate approach is interposition, wherein extension code is interposed at well-defined system interfaces SLIC. This approach requires no modifications to existing operating system code - instead, each call to an operation in the interface is intercepted and routed to the extension code.

System-call interposition, which involves interposing extension code at the application to operating system boundary, is one of the well-researched interposition approaches, discretionary access control, file system encryption, stackable file systems, sandboxing, intrusion prevention and detection, performance isolation, checkpointing, process migration as it offers several advantages over other forms of interposition. We propose a new framework that is inspired by the network packet-filtering model to develop system-call interposition based extensions called hybrid interposition. The key contributions of our approach over the current state of the art in system call interposition are: (a) A high level rule-based language called Behavior Monitoring Specification Language (BMSL) which provides an expressive, efficient mechanism to code extensions and (b) a hybrid interposition mechanism which intercepts system calls inside the kernel and processes them in user space.

William Hsu

Learning class-level generalizations from objective data is an important problem in link analysis. In this talk, I will describe several applications in data mining: collaborative recommendation for personalized Grid computing in the domain of bioinformatics; market basket analysis; and spatial data cleaning in modeling of water use systems. These domains have some properties and problems in common, namely: the need to find good relational representations; the problems of linkage and autocorrelation in learning probabilistic relational models (PRMs) from data; and the question of how to identify relevant variables in constructing class-level generalizations of relational data.

Collaborative recommendation in information retrieval (IR) is the problem of analyzing the content of an IR system and actions of its users to predict additional topics or products a new user may find useful. In our application test bed, the DESCRIBER project, we seek to map workflows (operational descriptions of experimental procedures) in computational genomics and proteomics, to object-oriented views, then generalize these to relational graphical models that can be used for link analysis, to recommend reusable workflow templates and components to researchers building new models. The challenge is to select or construct dynamic relational attributes for generative and discriminative graphical models.

Our approach extends some current techniques in collaborative recommendation by market basket analysis for commercial applications such as cross-selling. I will conclude with a brief discussion of a third related application of learning in graphical relational models that exploits dynamic relational attributes: the problem of "identity uncertainty" in spatial water resource localization from coalesced databases.

Chris Harding

Here is the abstract in pdf format.

The modeling (or interpretation) of geological structures from raw data is central to many geoscientific tasks in academia or industry. The advent of immersive displays and other VR technologies in the last years have made it easier for interpreters to interact directly with geoscientific data and ultimately to translate the models developed in their ¡°mind¡¯s eye¡± into a computer model. While geoscience data is now routinely visualized in 3D environments (for example in virtual theaters used by all major oil companies), the addition of touch and sound is still a research topic. The talk focuses on a prototype desktop virtual environment for the investigation of surface meshes - not only via 3D stereo but also with a point-haptic device called the Phantom and with a real-time audio stream. The haptic device is used to digitize line features, such as fault lines, directly on the surface while simultaneously feeling its morphology. The audio stream is a MIDI-interpretation of a (possibly secondary) data attribute such as terrain slope or gravity and useful in augmenting the interaction with the surface data.. The prototype system was primarily used to interpret faults surfaces and other tectonic structures along Mid-ocean ridge in the North Atlantic.

Rajat Todi is pursuing his Ph.D. degree from Department of Computer Science, Iowa State University under the supervision of Dr. Gurpur Prabhu. He is working as a Performance Engineer in the Systems VLSI Technology Division at Hewlett Packard (HP). At HP, Rajat Todi is in a team that is responsible for evaluating future Itanium Processor Family based microprocessors. Before joining HP, he worked in Ames Laboratory (USDOE), as a research assistant to Dr. John Gustafson for five years. He received his B.E. (honors) degree in Computer Science and M.S. (honors) degree in Mathematics from Birla Institute of Technology and Science, Pilani, India. He is a member of IEEE, IEEE-Computer Society, and ACM.

Visit Rajat Todi's homepage here.

BACK

Joseph P. Bigus

Dr. Joseph P. Bigus is a Senior Technical Staff Member at the IBM T. J. Watson Research Center, where he is the project leader on the Agent Building and Learning Environment (ABLE) research project. He is a member of the IBM Academy of Technology and is an IBM Master Inventor, with over 20 US patents. Joe was an architect of the IBM Neural Network Utility and Intelligent Miner for Data products. He received his M.S. (1988) and Ph.D. degrees (1993) in Computer Science from Lehigh University and a B.S. in Computer Science from Villanova University (1986). He has written two books: Data Mining with Neural Networks (McGraw-Hill) and Constructing Intelligent Agents using Java, 2nd ed. (Wiley). Dr. Bigus's current research interests include learning algorithms and intelligent agents, as well as multi-agent teams and their applications to autonomic computing, problem determination, data mining, and decision support. He is a member of the IEEE Computer Society and the International Neural Network Society.

Visit Joseph P. Bigus's homepage here.

BACK

Ramesh Jain

Ramesh Jain is an entrepreneur, educator, and a researcher.

He is the Rhesa "Ray" S Farmer Distinguished Chair in Embedded Experiential Systems and Georgia Research Alliance Eminent Scholar in School of Electrical and Computer Engineering and College of Computing at Georgia Institute of Technology. Ramesh is a world-renowned pioneer in multimedia information systems, image databases, machine vision, and intelligent systems. While professor of computer science and engineering at the University of Michigan, Ann Arbor and the University of California, San Diego, he founded and directed artificial intelligence and multimedia information systems labs. Jain was also the founding Editor-in-Chief of IEEE MultiMedia magazine and serves on the editorial boards of several magazines in multimedia, business and image and vision processing. He has co-authored more than 250 research papers in well-respected journals and conference proceedings. He has co-authored and co-edited several books, including Machine Vision, a textbook used at several universities.

He founded three companies, managed them in initial stages, and then turned them over to professional management. Most recently, he was the co-founder, CEO, and CTO of PRAJA Inc located in San Diego. PRAJA was acquired by Tibco. Prior to PRAJA, he was the founding CEO and Chairman of Virage (NASD: VRGE), a San Mateo-based company developing systems for media management solutions and visual information management. He was also the Founder and Chairman of ImageWare Inc. that provided solutions for surface modeling, reverse engineering rapid prototyping, and inspection. ImageWare was acquired by SDRC. He serves as advisors to several companies.

Visit Ramesh Jain's homepage here.

BACK

Joseph Halpern

Joseph Y. Halpern received a B.Sc. in mathematics from the University of Toronto in 1975 and a Ph.D. in mathematics from Harvard in 1981. In between, he spent two years as the head of the Mathematics Department at Bawku Secondary School, in Ghana. After a year as a visiting scientist at MIT, he joined the IBM Almaden Research Center in 1982, where he remained until 1996. He served as manager of the Mathematics and Related Computer Science Department at IBM from 1988-1990 and was a consulting professor at Stanford from 1984-1996. In 1996, he moved to Cornell University, where he is a professor in Computer Science.

His major research interests are in reasoning about knowledge and uncertainty, qualitative reasoning, belief revision, (fault-tolerant) distributed computation, game theory, decision theory, and security. Together with his former student, Yoram Moses, he pioneered the approach of applying reasoning about knowledge to analyzing distributed protocols and multi-agent systems. He has coauthored 5 patents, two books ("Reasoning About Knowledge" and "Reasoning about Uncertainty"), and well over 100 technical publications.

Halpern was program chairman and organizer of the first conference on Theoretical Aspects of Reasoning about Knowledge, program chairman of the fifth ACM Symposium on Principles of Distributed Computing, the 23rd ACM Symposium on Theory of Computing, and the 16th IEEE Symposium on Logic in Computer Science. He received the Publishers' Prize for Best Paper at at the International Joint Conference on Artificial Intelligence in 1985 (joint with Ronald Fagin) and in 1989, the 1997 Godel Prize (joint with Yoram Moses), and two IBM Outstanding Innovation Awards. He is a Fellow of the American Association of Artificial Intelligence and the Association for Computing Machinery, and in 2001-02 was the recipient of a Guggenheim and a Fulbright Fellowship. He was editor-in-chief of Journal of the ACM, and currently serves on the editorial board of Journal of Logic and Computation, Chicago Journal of Theoretical Computer Science, and Artificial Intelligence

Visit Joseph Halpern's homepage here.

BACK

Joseph Halpern

Joseph Y. Halpern received a B.Sc. in mathematics from the University of Toronto in 1975 and a Ph.D. in mathematics from Harvard in 1981. In between, he spent two years as the head of the Mathematics Department at Bawku Secondary School, in Ghana. After a year as a visiting scientist at MIT, he joined the IBM Almaden Research Center in 1982, where he remained until 1996. He served as manager of the Mathematics and Related Computer Science Department at IBM from 1988-1990 and was a consulting professor at Stanford from 1984-1996. In 1996, he moved to Cornell University, where he is a professor in Computer Science.

His major research interests are in reasoning about knowledge and uncertainty, qualitative reasoning, belief revision, (fault-tolerant) distributed computation, game theory, decision theory, and security. Together with his former student, Yoram Moses, he pioneered the approach of applying reasoning about knowledge to analyzing distributed protocols and multi-agent systems. He has coauthored 5 patents, two books ("Reasoning About Knowledge" and "Reasoning about Uncertainty"), and well over 100 technical publications.

Halpern was program chairman and organizer of the first conference on Theoretical Aspects of Reasoning about Knowledge, program chairman of the fifth ACM Symposium on Principles of Distributed Computing, the 23rd ACM Symposium on Theory of Computing, and the 16th IEEE Symposium on Logic in Computer Science. He received the Publishers' Prize for Best Paper at at the International Joint Conference on Artificial Intelligence in 1985 (joint with Ronald Fagin) and in 1989, the 1997 Godel Prize (joint with Yoram Moses), and two IBM Outstanding Innovation Awards. He is a Fellow of the American Association of Artificial Intelligence and the Association for Computing Machinery, and in 2001-02 was the recipient of a Guggenheim and a Fulbright Fellowship. He was editor-in-chief of Journal of the ACM, and currently serves on the editorial board of Journal of Logic and Computation, Chicago Journal of Theoretical Computer Science, and Artificial Intelligence

Visit Joseph Halpern's homepage here.

BACK

James Hendler

James Hendler is one of the originators of the Semantic Web. He is a Professor at the University of Maryland where he is the Director for Semantic Web and Agent Technology at the Maryland Information and Network Dynamics Laboratory. He has joint appointments in the Department of Computer Science, the Institute for Advanced Computer Studies and the Institute for Systems Research, and he is also an affiliate of the Electrical Engineering Department. He has authored close to 150 technical papers in areas including artificial intelligence, robotics, agent-based computing and high performance processing. Hendler was the recipient of a 1995 Fulbright Foundation Fellowship, is a Fellow of the American Association for Artificial Intelligence, and was decorated with the US Air Force Exceptional Civilian Service Medal in 2002. He is also the former Chief Scientist for Information Systems at the US Defense Advanced Research Projects Agency (DARPA), an advisor to NASA's Earth Science Activity and chairs the W3C's Web Ontology Working Group.

Visit James Hendler's homepage here.

BACK

Pierre Baldi

Professor Baldi received his Ph.D. in Mathematics from Caltech in 1986. He is the Director of the Institute for Genomics and Bioinformatics and holds joint appointments in the Department of Information and Computer Science and the College of Medicine's Department of Biological Chemistry at the University of California at Irvine. Baldi has authored several books including 'The Shattered-Self: The End of Natural Evolution' and 'Bioinformatics - The Machine Learning Approach' (with Soren Brunak), and 'Modeling the Internet and the Web - Probabilistic Methods and Algorithms' (with Paolo Frasconi and Padhraic Smyth) and'DNA Microarrays and Gene Expression' (with Wesley Hatfield). Baldi's research focuses on several areas of data mining, machine learning, bioinformatics and communication networks. His recent work has focused on computational methods for understanding and predicting protein structures, analyzing and modeling gene expression data and regulatory networks, analyzing and designing communication networks (Internet, Ultra Wide Band Radio) and quantifying information.

Visit Pierre Baldi's homepage here.

BACK

Michael Pazzani

Visit Michael Pazzani's homepage here.

BACK

Stephen Smale

Professor Smale received his PhD from the University of Michigan in 1957, and within four years became a full Professor at Columbia University. He became Professor at the University of California, Berkeley in 1964 and Professor Emeritus in 1994. Professor Smale became a Distinguished University Professor at the City University of Hong Kong in 1995. He is also a on the faculty of Toyota Technological Institute at Chicago.

He has made significant contributions in the fields of dynamic systems, geometry, econometrics, operational research, topology, and theoretical computer science. These contributions have resulted in a number of academic awards and achievements including his holding of the Alfred Sloan Research Fellowship from 1960-62. In 1966 he won a Fields Medal - an international medal awarded once every four years for outstanding discoveries in mathematics. Other awards he received include the 1965 Veblen Prize for Geometry, awarded every five years by the American Mathematical Society; in 1988 the Chauvenet Prize by the Mathematical Association of America; and in 1989 the Von Neumann Award by the Society for Industrial and Applied Mathematics.

Professor Smale is a member of both the National Academy of Sciences and the American Academy of Arts and Sciences. He is recognized internationally in many fields of Mathematics, and has been invited as a Visiting Professor to College de France, Paris (Spring 1962), University of Paris, Orsay (1972-73), Yale University (Fall 1974), and Columbia University (Fall 1987).

Visit Stephen Smale's homepage here.

BACK

Thad Starner

Thad Starner is a wearable computing pioneer, completing one of the first PhDs addressing the subject while at the MIT. Thad has been using a wearable computer as an integral part of his everyday life since 1993. Starner is an assistant professor in the College of Computing at Georgia Tech and has authored over 70 scientific publications in computing. Thad's current work researches the use of computational agents for everyday-use wearable computers as a segue to artificial intelligence.

Visit Thad Starner's homepage here.

BACK

Raghu Ramakrishnan

Raghu Ramakrishnan received his B.Tech from the Indian Institute of Technology at Madras and his Ph.D. in Computer Science from the University of Texas at Austin. He is Professor of Computer Sciences and Vilas Associate at the University of Wisconsin-Madison, and Chairman and CTO of QUIQ, a company that is pioneering collaborative customer support. His research is in the area of database systems, in particular databases and information retrieval, data mining, and query optimization. He is editor-in-chief of the Journal of Data Mining and Knowledge Discovery, and co-director of the UW-Madison Data Mining Institute. Dr. Ramakrishnan is a Fellow of the Association for Computing Machinery (ACM), and has received several awards, including a Packard Foundation Fellowship and an ACM SIGMOD Contributions Award. He has authored over 100 technical papers and written the widely-used text "Database Management Systems" (WCB/McGraw-Hill), now in its second edition (with J. Gehrke).

Visit Raghu Ramakrishnan's homepage here.

BACK

Satinder Singh

Satinder Singh is an Associate Professor of Electrical Engineering and Computer Science at the University of Michigan, Ann Arbor. Prior to this he has been a Principal Research Scientist at AT&T Labs, an Assistant Professor of Computer Science at the University of Colorado, Boulder, and a Postdoctoral Fellow at MIT's Brain and Cognitive Science department. His research focus is on developing the theory, algorithms and practice of building agents that can learn from interaction in complex, dynamic, and uncertain environments, including environments with other agents in them. His main contributions have been to the areas of reinforcement learning and more recently multi-agent learning. He edited a special issue on reinforcement learning for the Machine Learning journal in 2002, has coauthored more than 60 refereed papers in journals and conferences and has served on many program committee's (AAAI, ICML, NIPS, UAI, COLT) and on journal editorial boards (Machine Learning, JAIR, JMLR).

Visit Satinder Singh's homepage here.

BACK

Rich Sutton

Rich Sutton received the B.A. degree in psychology from Stanford University in 1978 and the M.S. and Ph.D. degrees in Computer Science from the University of Massachusetts in 1980 and 1984. He worked for nine years at GTE Laboratories as principal investigator of their connectionist machine learning project, and for three years at the University of Massachusetts as a research scientist in the computer science department. In 1998-2002, Rich worked at AT&T Labs in Florham Park, New Jersey, and in 2003 he became professor of Computing Science at the University of Alberta. He is also a fellow of the American Association for Artificial Intelligence.

Rich's research interests center on the learning problems facing a decision-maker interacting with its environment, which he sees as central to artificial intelligence. He is the author of the original paper on temporal-difference learning and, with Andrew Barto, of the textbook Reinforcement Learning: An Introduction. He is also interested in animal learning psychology, in connectionist networks, and generally in systems that continually improve their understanding and representations of their world.

Visit Rich Sutton's homepage here.

BACK

Prem Uppuluri

Prem Uppuluri has a BE in Computer Science from Osmania University India, M.S from Iowa State University and Ph.D from the State Univ. of New York at Stony Brook (2003). He is currently an Asst. Professor in Computer Science at the University of Missouri, Kansas City. His research includes computer system security (intrusion prevention, secure operating systems), network management (protocol monitoring and fault injection) and operating systems (interposing extension code and file system).

BACK

William Hsu

William Hsu received a Ph.D. in Computer Science from the University of Illinois at Urbana-Champaign in 1998 and served as a research scientist in the Automated Learning Group at the National Center for Supercomputing Applications before joining the faculty of Computing and Information Sciences at Kansas State University in 1999. His research interests are in the areas of machine learning and probabilistic reasoning, with applications to data mining, decision support, autonomous robotics, multi-agent systems, and computational genomics.

Visit William Hsu's homepage here.

BACK

Chris Harding

Chris Harding received a Masters Degree in 1993 in Geology from the Free University in Berlin, Germany, where he specialized in mathematical geology such as geostatistics and Geographic Information Systems (GIS). He worked as a programmer for mine planning and geotechnical applications and received a Ph.D. in Geology from the University of Houston in 2001 where he worked on multi- sensory Virtual Environments for petroleum exploration data. He worked for ExxonMobil and Shell on the application of audio and force-feedback (haptics) to improve seismic interpretation. He is currently employed as Assistant professor at the geoscience department at Iowa State University and is also part of its Human-Computer-interaction (HCI) program which is centered at the Virtual Reality Applications Center (VRAC).

BACK

---

Archives

• Spring 2003
• Fall 2002
• Spring 2002
• Fall 2001
• Spring 2001
• Fall 2000
• Spring 2000
• Fall 1999
• Spring 1999
• Fall 1998
• Summer 1998
• Spring 1998
• Fall 1997
• Spring 1997

(Top)

---

Contacts

Thank you for visiting this page. Please send your suggestions and comments to one of us in the Computer Science colloquium committee.

• Soma Chaudhuri . chaudhur@cs.iastate.edu
• Dalei Li. daleili@cs.iastate.edu

(Top)

---