Shell Distinguished Lectures1998-1999 AbstractsScalable Parallel Computers, Symmetric Multiprocessors and ClustersMarc Snir,IBM TJ Watson Research Center 4:00pm, Monday January 25, 1999
AbstractThe last few years have seen significant growth in two types of parallel computer architectures: cache coherent, distributed shared memory (NUMA) systems, controlled by one operating system image; and integrated clusters, controlled by multiple operating system images. These two types of systems co-exist in the market place because they answer different needs. However, recent research and emerging products indicate the possibility for hybrid systems that combine the best of both worlds. We shall discuss in our talk the requirements that drive these two types of high end server architectures and the technologies that may lead to their eventual unification.
BiographyMarc Snir is a senior manager at the IBM T. J. Watson Research Center, where he leads research on scalable parallel software and on scalable parallel architectures. The group led by him was responsible for major contributions to the SP scalable parallel computers: architecture, parallel operating environment, message-passing libraries, tools, parallel file system, parallel algorithms and applications.He received a Ph.D. in Mathematics from the Hebrew University of Jerusalem in 1979. He worked at NYU on the NYU Ultracomputer project from 1980-1982, and worked at the Hebrew University of Jerusalem from 1982-1986. More recently, he has contributed to the design of High Performance Fortran and the Message Passing Interface. He has published numerous papers and given many presentations on computational complexity, parallel algorithms, parallel architectures, interconnection networks, and parallel programming environments. Dr. Snir is ACM Fellow, IEEE Fellow, and member of the IBM Academy of Technology. He is on the editorial board of Parallel Processing Letters and Mathematical Systems Theory.
Randomized Path Planning: Algorithms, Analysis, and ApplicationsJean-Claude Latombe,Stanford University 4:00pm, Monday February 8, 1999
AbstractThe inherent complexity of path planning with many degrees of freedom has motivated the investigation of planners with only partial completeness guaranties. In this context, randomized planners are particularly promising: (1) they make it possible to trade full completeness against a slightly weaker (but still useful) form of completeness -- probabilistic completeness -- while achieving major gains in computational time; (2) they also yield new, more intuitive characterization of planning complexity than traditional planners. The randomized planners considered in this talk operate by randomly sampling the robot's configuration space and making simple connections among the sampled points, yielding a data structure called a "probabilistic roadmap". In 1995, we introduced the geometric notion of "epsilon-goodness" to formally analyze and explain the empirical success of this scheme. A shortcoming of this notion led us to introduce a slightly stronger notion, "expansiveness", to characterize configuration spaces without narrow passages. In an expansive configuration space, the randomized planning scheme solves path planning problems correctly and efficiently, with high probability. The expansiveness assumption eliminates some important spaces, but it yields a new sampling strategy to deal with narrow passages. This talk will present these successive notions and their relation to planning efficiency. It will also show results on practical application problems: disassembly of complex mechanical products, animation of digital actors, and analysis of molecular binding motions. It will discuss recent extensions to dynamic planning with moving obstacles and optimal trajectory planning.
BiographyJean-Claude Latombe is professor of Computer Science at Stanford University. He is also the current chair of the Computer Science Department. He received a Ph.D. degree in Computer Science from the National Polytechnic Institute of Grenoble, France, in 1977. His current research investigates algorithms and system architectures for geometric and physical reasoning, with the goal to create autonomous agents that sense, plan, and act in real and/or virtual worlds. He is particularly interested in robot-assisted surgery, design and manufacturing, digital actors, drug design, and active observation of remote physical environments over the Internet.
AI Challenges and Opportunities for the 21st CenturyDavid Waltz,NEC Research Institute 4:00pm, Monday March 8, 1999
AbstractAI has been successful in fielding applications that mine large databases for facts and patterns, that plan for complex logistics problems, that allow programs to accept speech inputs, that help physicians diagnose diseases and monitor patients, that play games (e.g. chess, bridge, poker) at human levels, etc. But there are still no computational examples of human-level performance on many human abilities (e.g. visual perception, deep natural language understanding, flexible reasoning). In fact, there is not even a consensus about the nature of the mechanisms that underlie such human abilities. But such abilities are important for future applications such as intelligent environments, instructable assistants, and text search systems that actually find meaningful answers to questions rather than merely returning a number of texts that contain the words of a query.This talk will argue that any system we would consider truly intelligence will depend critically on the ability to separate the important from the unimportant. AI has for the most part focussed on logic and reasoning in artificial situations where only relevant variables and operators are specified, and has paid insufficient attention to processes of reducing the richness and disorganization of the real world to a form where logical reasoning can be applied. This talk will discuss the role of importance in intelligence, provide some positive examples of research that makes use of importance judgements, and offer suggestions for new mechanisms, architectures, applications, and research directions for AI.
BiographyDavid Waltz has been Vice President of the Computer Science Research Division of NEC Research Institute in Princeton, NJ. He is currently President of the AAAI and is a Fellow of the AAAI, a Fellow of the ACM, and a Senior Member of IEEE, and a former chair of ACM SIGART. Before moving to NEC Research, Dr. Waltz directed the data mining and text retrieval effort at Thinking Machines Corp. for nine years, and was at the same time on the Brandeis University faculty; he still holds an Adjunct Professor position at Brandeis. From 1973 to 1984 he was on the faculty at the University of Illinois at Urbana-Champaign. Dr. Waltz received all his degrees from MIT. His research interests have included constraint propagation, computer vision, massively parallel systems for both relational and text databases, memory-based and case-based reasoning systems and their applications, protein structure prediction using hybrid neural net and memory-based methods, and connectionist models for natural language processing.
Wearing it Out: First Steps Toward Mobile Augmented Reality SystemsSteven Feiner,Columbia University 4:00pm, Wednesday March 31, 1999
AbstractAs computers grow ever smaller and faster, the option of wearing them, rather than carrying or sitting in front of them, is rapidly becoming a reality. Redesigning computers and computing for wearability is a difficult, multifaceted problem. Not only must wearable machines literally fit our bodies in a way previously required only of interaction and display devices, but user interfaces themselves must also change. Approaches and metaphors that work well within the part-time, stationary microcosm of the desk-top may be far less effective when applied to the full-time, mobile environment.One especially promising approach for wearable user interfaces is "augmented reality". This alternative form of virtual reality augments the physical world with additional information, rather than replacing it. As pioneered by Ivan Sutherland, this can be accomplished through the use of head-tracked "see-through" displays that enrich the user's experience of the surrounding world by overlaying it with graphics, text, speech, sound, and other media. This talk provides an overview of the work being done by Columbia's Computer Graphics and User Interfaces Laboratory to address fundamental design issues for mobile augmented reality. Such systems must function indoors and outdoors, both stand-alone and together with many other displays, devices, and users, into and out of whose presence we move. We will describe our distributed software infrastructure, which supports collaborative interaction across heterogeneous architectures. Building on this infrastructure, we are developing research prototypes that use a mix of displays, trackers, and interaction devices to create mobile augmented reality systems. One example that we will present combines GPS position tracking and orientation tracking with head-worn and hand-held displays to allow a user to explore an unfamiliar urban environment. Another supports collaboration by embedding the 2D GUIs of stationary, laptop, and hand-held computers within a shared, overlaid, 3D virtual surround.
BiographyDr. Steven Feiner is an Associate Professor of Computer Science at Columbia University, where he directs the Computer Graphics and User Interfaces Laboratory. He received a Ph.D. in Computer Science from Brown University. His research interests include user interfaces, virtual environments and augmented reality, knowledge-based design of graphics and multimedia, information visualization, wearable computing, image synthesis, and hypermedia.Prof. Feiner is coauthor of Computer Graphics: Principles and Practice (Addison-Wesley, 1990) and of Introduction to Computer Graphics (Addison-Wesley, 1993). He is an associate editor of ACM Transactions on Graphics and a member of the editorial board of IEEE Transactions on Visualization and Computer Graphics. Prof. Feiner serves on the executive boards of the IEEE Technical Committee on Computer Graphics and the IEEE Task Force on Human-Centered Information Systems. In 1991 he received an Office of Naval Research Young Investigator Award.
Dynamic Power ManagementGiovanni De Micheli,Stanford University 10:30am, Tuesday June 1, 1999
AbstractDynamic Power Management is a design methodology aiming at controlling performance and power levels of digital circuits and systems, with the goal of extending the autonomous operation time of battery-powered systems, providing graceful performance degradation when supply energy is limited, and adapting power dissipation to satisfy environmental constraints.This talk we survey dynamic power management applied at the sytesm level. We analyze first idleness detection and shutdown mechanisms for idle hardware resources. We review industrial standards for operating system-based power management, such as the Advanced Configuration and Power Interface (ACPI) standard proposed by Intel, Microsoft and Toshiba. Next, we review system-level modeling techniques, and describe stochastic models for the power/performance behavior of systems. We analyze different modeling assumptions and we discuss their validity. Last, we describe a method for determining optimum policies and validation methods, via simulation at different abstraction levels, for power managed systems.
BiographyGiovanni De Micheli is Professor of Electrical Engineering, and by courtesy, of Computer Science at Stanford University. Previously he held positions at the IBM T.J. Watson Research Center, Yorktown Heights, New York, at the Department of Electronics of the Politecnico di Milano, Italy and at Harris Semiconductor, Melbourne, Florida. He holds a Nuclear Engineer degree (Politecnico di Milano, 1979), and a M.S. and Ph.D. degree in Electrical Engineering and Computer Science (University of California at Berkeley, 1980 and 1983).His research interests include several aspects of the computer-aided design of integrated circuits and systems, with particular emphasis on synthesis, system-level design and hardware/software co-design. He is author of: Synthesis and Optimization of Digital Circuits, McGraw-Hill, 1994; co-author of: Dynamic Power management: Circuit Techniques and CAD Tools, Kluwer 1998, and of High-level Synthesis of ASICs under Timing and Synchronization Constraints; and co-editor of: Hardware/Software Co-design, Kluwer, 1995 and of Design Systems for VLSI Circuits: Logic Synthesis and Silicon Compilation, Martinus Nijhoff Publishers, 1986. He was also co-director of the NATO Advanced Study Institutes on Hardware/Software Co-design, held in Tremezzo, Italy, 1995 and on Logic Synthesis and Silicon Compilation, held in L'Aquila, Italy, 1986. Dr. De Micheli is a Fellow of IEEE. He was granted a Presidential Young Investigator award in 1988. He received the 1987 IEEE Transactions on CAD/ICAS Best Paper Award and two Best Paper Awards at the Design Automation Conference, in 1983 and in 1993. He is Vice President (for publications) of the IEEE CAS Society. He is the Editor of the IEEE Transactions on CAD/ICAS. He was the Program Chair (for Design Tools) of the Design Automation Conference in 1996 and 1997, and he is currently Vice General Chair of DAC. He was Program and General Chair of International Conference on Computer Design (ICCD) in 1988 and 1989 respectively. |
| ||||
| ||||