2004 PSP Annual Meeting Home     HOME    |    AGENDA    |    REGISTRATION    |    FAQ   
2004 PSP Annual Meeting Home

2004 Private Sector Program Annual Meeting Abstracts

NCSA Resources to Support PSP Partner Effectiveness
John Towns
Monday, April 26
2:00 pm
Holiday Inn

Various resources exist at NCSA in it's high end computing environment that are currently available to NCSA's Private Sector Partners. In this presentation we will discuss the available high end computing and data resources that exist along with various Independent Software Vendor (ISV) applications that are available. Matching up the resources to needs within the various PSP partners is a non-trivial process and it is desired that this not only be a presentation of available capabilities but also a discussion of how NCSA can configure and present these in ways amenable to use by the PSP partners.

 

 

HDF5
Mike Folk
Monday, April 26
6:30 pm
5602 Beckman Institute

Collecting and storing data from test systems and platforms has historically been reduced to unique in-house implementations. Man-months to man-years have been expended to create and develop these site and system specific storage file formats.

NCSA's HDF5 data management system has the functionality and performance needed to capture, store and retrieve flight test data—it is comprehensive, scalable, flexible, and fast. Using HDF5 as the initial container for test data and replacing all the subsequent intermediate formats will yield cost savings for both non-recurring and recurring items associated with test data collection and processing. Ultimately, this work can lead to standardized formats, software, and tools benefiting a wide range of test systems in many industries.

The focus in the first year of this project will be on variable length array storage of test data in HDF5.

 

 

EMO - Evolutionary Multi-objective Optimization
Michael Welge, Loretta Auvil, David Clutter
Monday, April 26
6:30 pm
1005 Beckman Institute

Decision-makers today need tools that can provide rapid guidance on the tradeoffs among numerous, often conflicting, objectives. For example, in the financial portfolio management domain, decision-makers must be able to rapidly identify the tradeoffs among risk, reward, transaction costs, etc. in potential portfolios. Presenting these tradeoffs in an intuitive, easily understood format enables more informed and accurate decision-making. These tradeoffs can be identified using the evolutionary multi-objective optimization (EMO) software developed in this project.

Discover has recently named evolutionary computation methods such as EMO one of the top 50 discoveries in science. They have been used for numerous practical applications including: designing jet engines, steam turbines, electrical power systems, microwave antennae, telecommunication networks, environmental remediation and monitoring, electrical circuits, etc.; fighting terrorism; operating chemical plants; scheduling factories; trading stocks; reconstructing criminal faces from witness recollection; calculating optimal interplanetary spacecraft trajectories; routing vehicles; placing communication satellites; choosing financial portfolios; and much, much more. Successes in these areas have been featured on ABC Nightline, in the New York Times, Wall Street Journal, Discover Magazine, and U.S. News and World Reports.

In the past, applications of EMO have required extensive expertise and programming skills to use the algorithms effectively. Recent advances in the EMO field now make it possible to implement these algorithms within interactive, collaborative decision environments so that the algorithms run completely "behind the scenes" without user intervention or knowledge. In a previous TRECC project, the EMO software was designed and tested on hypothetical cases. The objective of this project was to complete the software interface, providing a complete prototype of the EMO software, and begin transferring the technology into practice.

EMO provides a tool for applying genetic algorithm (GA) optimization in a general framework. It includes single objective genetic algorithms as well as multi-objective genetic algorithms using an approach called NSGA-II. EMO is built on the Automated Learning Group's D2K and D2K-SL software, providing all the sophistication and power of the D2K infrastructure. There are D2K modules for each step of the genetic algorithm process: generating populations, evaluating populations, evaluating constraints, crossover techniques, mutation techniques, and selection techniques.

The EMO technology holds great promise for helping decision makers to identify tradeoffs among complex objectives. The prototype interface for EMO has been completed and technology transfer is well under way. EMO is in the process of being licensed to Moire Inc. for use in optimizing long-term groundwater monitoring at subsurface waste sites.

 

 

River Web
Vernon Burton
Monday, April 26
6:30 pm
3269 Beckman Institute

River Web is a collaborative, Web-based, multimedia education project about the Mississippi River and its impact on the lives of people over time. The architecture of River Web is adaptable for any community across the U.S. or world, as River Web can include many "landing sites" worthy of investigation. More than a program, River Web is a learning environment, giving students access to original documents and artifacts, enabling them to "do" research instead of simply reading about it, contributing to their sense of genuine discovery. We know of no other model with the potential of River Web to transform the use of technology in college history, humanities, and social studies classrooms. River Web solves two problems that colleges across the nation are striving to achieve: integrating information technology into inquiry-based education; management of the huge amount of potentially relevant information on the Internet. River Web provides a set of primary source materials that follow selected criteria, allowing students to explore a topic along a cohesive pathway.

River Web has been used regularly in the classroom since 1998; we have a prototype that has been validated and is known to work. We now have two objectives: 1) we need to expand its Digital Library component (both archives and tools) and create an interactive, online classroom with learning modules; 2) we want to engage and support faculty from a variety of 2-year and 4-year institutions, including under-resourced schools and those serving minorities (Historically Black Colleges and Universities, Hispanic Serving Institutions, and Tribal Colleges and Universities) in the process of integrating River Web into their courses. An outside evaluator will conduct systematic quantitative and qualitative evaluation studies to document the achievement of these objectives and to provide a scientific basis for River Web as a national model for technology integration in college history, humanities, and social studies.

River Web provides high performance computing, internet software development, multimedia integration, and graphic design, all to further educational benefits that are far-reaching and significant. This project will increase the knowledge base of undergraduate education and enhance student learning because of the inquiry-based environment it creates and facilitates. River Web will enhance the content and pedagogical preparation of faculty because its design is easily extensible to other geographic regions and to a broad range of learning opportunities.

 

 

MAIDS: Mining Alarming Incidents in Data Streams
Dora Cai, Michael Welge, Jiawei Han, David Clutter, Grep Pape, Loretta Auvil
Monday, April 26
6:30 pm
1005 Beckman Institute

Data streams are transient data that pass through the system continuously and in huge volumes. There are many applications that require handling data in the form of a stream, such as sensor data, network traffic flow, time-series data, stock exchange data, telecommunications, Web click streams, weather or environment monitoring, and so on. Data streams are different from the finite, static data sets stored in flat files or in database systems; they are in high volume, potentially infinite, dynamically changing, and require fast response. These unique characteristics make data stream analysis a great challenge.

MAIDS, Mining Alarming Incidents from Data Streams, is a project supported by U.S. Office of Naval Research and National Science Foundation, jointly developed by Automated Learning Group, NCSA and the Data Mining group (DAIS), Department of Computer Science, the University of Illinois at Urbana-Champaign. By integration of our most recent research results published in VLDB'2002, VLDB'2003, KDD'2003, etc, we have successfully developed the MAIDS system within the D2K (Data to Knowledge, developed by Automated Learning Group, NCSA) data mining framework with the following distinct features:

  1. A tilted time window framework and multi-resolution model,
  2. A stream "data cube" for multi-dimensional analysis,
  3. Online stream classification,
  4. Online frequent pattern mining,
  5. Online clustering of data streams,
  6. Stream mining visualization.

These stream data mining functions, working together, can successfully mine alarming incidents from data streams on the fly.

The system will be demonstrated using the stream data generated from a network logging process. The demo will be run continuously to show the activities of stream data mining. Users are welcome to interact with the system anytime and explore the detected network intrusions that are visualized in various charts and graphs.

 

 

Textmining - Information Extraction and Visualization of Streaming Text
Duane Searsmith
Monday, April 26
6:30 pm
1005 Beckman Institute

Using the T2K module suite in the D2K framework we will demonstrate information extraction of names, places and organizations from real-time news feeds. The information extracted will be visualized as a dynamic graph showing prominent items of interest and their relationships. We will also demonstrate the ease with which D2K/T2K allows integration with existing external libraries—for example, University of Sheffield's Generalized Architecture for Text Engineering (GATE).

In addition, we will also demonstrate a non-parametric pattern mining approach to building hierarchical topic trees. A topic directory is a hierarchical document tree or graph structure such that each node has a topic label and corresponding documents, and the topic of a higher node conceptually covers its children nodes. Such structures are used to efficiently discover, describe, and organize topical content in large collections of documents.

Text clustering and classification demonstrations will also be available.

D2K—Data to Knowledge—is a rapid, flexible data mining and machine learning system that integrates analytical data mining methods for prediction, discovery, and deviation detection, with data and information visualization tools. T2K is a suite of modules in the D2K framework for use in text mining, analysis, and visualization.

 

 

VisIT
Dan Kauwell
Monday, April 26
6:30 pm
5269 Beckman Institute

More than half of business professionals spend over two hours a day searching for information they need to perform their jobs. When looking for information, 79% of workers seek free information on the Internet. Due to the difficulties of finding needed information on the Internet, U.S. companies are spending $107 billion a year paying employees to search for external information. Even more problematic is the lack of tools for analyzing and sharing information once it is found. VisIT is an information retrieval and knowledge management environment based on a Web services architecture that is addressing these problems. VisIT has numerous applications in research, competitive intelligence and B2B purchasing.

 

 

Robot Control Using Arm Gestures
Rob Kooper
Monday, April 26
6:30 pm
1005 Beckman Institute

We present a system for gesture recognition using multiple orientation sensors. We focus specifically on the problem of controlling Unmanned Aerial Vehicles (UAVs) in the presence of manned aircrafts on an aircraft deck. Our goal was to design a UAV control with the same gesture signals as used by current flight directors for controlling manned vehicles. Given the fact that such a system has to operate 24 hours a day in a noisy and harsh environment, for example, on a Navy carrier deck, our approach to this problem is based on arm gesture recognition rather than on speech recognition. We have explored multiple approaches to arm gesture recognition, and investigated real-time and system design issues for a particular choice of active sensors. We describe several theoretical and experimental issues related to a design of a real-time gesture recognition system using the IS-300 Pro Precision Motion Tracker by InterSense. Our work consists of (1) analyzing several gesture recognition approaches leading to a selection of an active sensor, (2) scrutinizing sensor data acquisition parameters and reported arm orientation measurements, (3) choosing the most optimal attachment and placement of sensors, (4) measuring repeatability of our experiments using Dynamic Time Warping (DTW) metric, and (5) designing model-based and template-based gesture classification algorithms and robot control mechanisms, where the robot represents an UAV surrogate in a laboratory environment.

 

 

Modeling of Spectral Imagery
Peter Bajcsy
Monday, April 26
6:30 pm
1005 Beckman Institute

This demonstration presents a novel approach to multi-sensor statistical modeling of bi-directional texture functions (BTF). Our proposed BTF modeling approach is based on (1) conducting an analytical study that relates a sensor resolution to the size and shape of elements forming material surface, (2) developing a robotic system for laboratory BTF data acquisition, (3) researching an application of the Johnson family of statistical probability distribution functions (PDF) to BTF modeling, (4) selecting optimal feature space for statistical BTF modeling, (5) building a database of parameters for the Johnson family of PDFs that after interpolations forms a high-dimensional statistical BTF model and (6) researching several statistical quality metrics that can be used for verification and validation of the obtained BTF models. The motivation for developing the proposed statistical BTF modeling approach comes from the facts that (a) analytical models have to incorporate randomness of outdoor scene clutter surfaces and (b) models have to be computationally feasible with respect to the complexity of modeled interactions between light and materials. The major advantages of our approach over other approaches are (a) the low computational requirements on BTF modeling (BTF model storage, fast BTF model-based generation), (b) flexibility of the Johnson family of PDFs to cover a wide range of PDF shapes and (c) applicability of the BTF model to a wide range of spectral sensors, e.g., color, multi-spectral or hyperspectral cameras. The prime applications for the proposed BTF model are multi-sensor automatic target recognition (ATR), and scene understanding and simulation.

 

 

Tiled Display Wall Demo
Dave Semeraro
Monday, April 26
6:30 pm
3514 Beckman Institute

The NCSA Visualization and Virtual Environments group will be demonstrating its scalable tiled display wall, one of the highest resolution displays in existence. The current configuration is a 40-projector system driven by a cluster of 40 PCs, resulting in one logical display with a resolution of 8,192 x 3,840 pixels. This work represents NCSAs contribution to the scalable tiled display wall community and is being documented as part of the Alliance Wall-in-a-box initiative. The applications that will be shown include high-resolution movie playback, large image exploration, and interactive 3D visualization.

 

 

CAVE Demonstrations
Dave Semeraro
Monday, April 26
6:30 pm
3510 Beckman Institute

At the NCSA CAVE we will be demonstrating our latest virtual reality applications, displayed on our new projectors with increased resolution and clarity. The NCSA virtual reality applications have been developed through collaborations with industry, local research scientists, NCSA Faculty Fellows, University of Illinois students, and VR programmers from around the Alliance. Visitors to the CAVE will have the opportunity to see our newest work, or some old classics that may be new to them.

 

 

NCSA VIAS
Alan Craig and Kalev Leetaru
Monday, April 26
6:30 pm
5602 Beckman Institute

The NCSA VIAS system automatically gathers internet information resources from electronic mailing lists, newsgroups, newsfeeds, and targeted Web crawls on specific fields of interest, then applies an assortment of algorithms to the resulting database to extract metadata such as company names, bibliographic references, and over 15 more metadata types. Additionally, the system identifies a number of document types, such as resumes, job descriptions, etc. The system then provides a Web-based interface to support various kinds of queries against that database of information. This is a fully automated system, requiring no human intervention. The system runs in a low cost cluster environment allowing it to be scaled depending on the needs of the user. The system is also available as an API making the VIAS functionality available to application developers.

 

 

The Collaborative Continuum
Jason Leigh
Monday, April 26
6:30 pm
3514 Beckman Institute

This is a research project to understand how distantly located collaborators work in display-rich environments and how these environments can solve some of the problems typically associated with distance collaboration. The Continnum combines a broad range of technologies including a shared touch screen, an AccessGrid multi-site audio/video conferencing system, a tiled display for high resolution data visualization, and the GeoWall—a 3D stereo visualization screen. Two such environments exist at the Electronic Visualization Laboratory (EVL) at the University of Illinois at Chicago, and a third at the Technology Research, Education and Commercialization Center (TRECC) in Dupage County, Illinois.

 

 

NCSA/UIUC Research Collaboration with the Cognitive Computation Group
Dan Roth and Dan Kauwell
Monday, April 26
6:30 pm
5269 Beckman Institute

NCSA partnerships often involve collaborations with distinguished professors on the UIUC campus. This demo discusses one such collaboration between Professor Dan Roth and NCSA's VisIT development group.

One facet of Professor Roth's work is the development of methods for information extraction and intelligent access to free-form textual information. These are methods, rooted in Machine Learning and Natural Language Processing, that attempt to access free form texts, as if they were stored in a database with a known schema. This effort requires identifying, to some extent, the "semantics" of the written text.

Specific projects in this direction include:

  1. context sensitive identification and classification of functional phrases, such as those that represent names of people, locations, organizations, chemicals, etc.
  2. identifying relations between phrases, and
  3. context sensitive identification of and adaptation to variations in written text, such as using different names for the same materials or locations.

For more on VisIT, see the abstract.

 

 

Knowledge Center
Tim Wentling
Monday, April 26
6:30 pm
1003 Beckman Institute

The Knowledge Center is a one-stop location containing key tools for building and maintaining geographically distributed communities of practice. The Centers consist of four primary areas: Knowledge Base, Collaboration Space, Knowledge Exchange, and e-Learning. Each of these areas features tools specific to the various functions of sharing and integrating knowledge. Knowledge Centers are capable of support multiple communities of practice with each community having its own collaboration space, knowledge exchange, and e-Learning tool sets. Among the tools available within the various Knowledge Center systems are:

  • Expert Networks
  • Community Calendars
  • Document Repositories
  • e-Learning Systems with effortless content sharing, integrated tools for collaboration, & dynamic publishing
  • Intellectual Property Support for e-Learning content
  • Learner Notebooks
  • Dynamic Message Boards
  • Auto-archiving Chat Rooms
  • Email list integration
  • Workshop Mgmt System with proposal submission, review, and scheduling features
  • Group, Role, & Permission Security Controls
  • Custom "Dashboard" Monitoring Service
  • Full System Search Integration
  • Multipoint Video Conferencing

The demonstration being conducted Monday evening will feature exhibitions of each of these features within our newest Knowledge Center spotlighting the e-Learning domain. If you would like to see a Knowledge Center in action or wish to know more about our research on e-Learning, knowledge management, and the influence of culture on each, please stop by the KLSG demonstration table.

 

 

SIFT - Security Incident Fusion Tools for Visualizing Security
Kiran Lakkaraju, Xiaoxin Yin, and Yifan Li
Monday, April 26
6:30 pm
4269 Beckman Institute

NCSA Security Research has developed a visualization framework for network traffic that can be used to highlight security events. We will demonstrate two tools (1) NVisionIP which visualizes security on systems and (2) VisFlowConnect which visualizes security via link relationships between systems.

 

 

Cluster Security
Forrest Xin Meng and Greg Koenig
Monday, April 26
6:30 pm
4269 Beckman Institute

As clusters become larger they become more powerful but also more difficult to monitor, especially for security. NCSA Security Research is developing a suite of security tools specifically designed for the cluster environment. We will demonstrate a Cluster Process Monitoring Tool that keeps track of all the processes on large clusters to identify malicious and suspicious events.

 

 

Debut of New Web Tool
Alan Craig and Kalev Leetaru
Tuesday, April 27
11:15 am

Due to the uniqueness of this technology, we are unable to provide an abstract prior to the annual meeting. Details will be available for meeting attendees only—at the time of presentation.

 

 

Data Mining
Michael Welge, Automated Learning Group
Tuesday, April 27
2:00 pm

This presentation will encompass the past, present, and future of data mining at NCSA. It will highlight the development of D2K Data-to-Knowledge framework, the role of D2K in next-generation research and development activities on-going at Social Security Administration, Office of Naval Research, and National Center for Secure Systems Research, and the transfer of the D2K technology to the commercial sector.

 

 

Research to Practice: Facilitating Knowledge Sharing through Technology
Tim Wentling
Tuesday, April 27
2:45 pm

Ninety percent of learned information within organizations is not shared. To compensate for this lack of knowledge transfer, organizations spend in excess of $20 billion on training and $13.4 billion on knowledge management. NCSA's Knowledge & Learning Systems Group is researching and blending the best practices from e-Learning and Knowledge Management in order to build its innovative knowledge sharing environment, the Knowledge Center. Knowledge Centers are integrated environments featuring dynamic knowledge respositories, ask-an-expert systems, e-Learning development and delivery systems, and a collection of collaboration tools to assist users in acquiring, sharing, and learning new knowledge either on their own or as part of an active community of practice.

This presentation provides an overview of some of the current and planned research that NCSA is pursuing in the arena of knowledge sharing and communities of practice. An overview of the NCSA Knowledge Center and how it is being deployed will be presented. This presentation also will preview upcoming changes in technologies and business practices that will affect training and knowledge management practices of organizations worldwide.

 

 

Robot Control Using Arm Gestures
Peter Bajcsy, Research Scientist, Automated Learning Group
Tuesday, April 27
4:00 pm

We present a system for gesture recognition using multiple orientation sensors. We focus specifically on the problem of controlling Unmanned Aerial Vehicles (UAVs) in the presence of manned aircrafts on an aircraft deck. Our goal was to design a UAV control with the same gesture signals as used by current flight directors for controlling manned vehicles. Given the fact that such a system has to operate 24 hours a day in a noisy and harsh environment, for example, on a Navy carrier deck, our approach to this problem is based on arm gesture recognition rather than on speech recognition. We have explored multiple approaches to arm gesture recognition, and investigated real-time and system design issues for a particular choice of active sensors. We describe several theoretical and experimental issues related to a design of a real-time gesture recognition system using the IS-300 Pro Precision Motion Tracker by InterSense. Our work consists of (1) analyzing several gesture recognition approaches leading to a selection of an active sensor, (2) scrutinizing sensor data acquisition parameters and reported arm orientation measurements, (3) choosing the most optimal attachment and placement of sensors, (4) measuring repeatability of our experiments using Dynamic Time Warping (DTW) metric, and (5) designing model-based and template-based gesture classification algorithms and robot control mechanisms, where the robot represents an UAV surrogate in a laboratory environment.

 

 

 

 National Center for Supercomputing Applications
605 East Springfield Avenue
Champaign, Illinois 61820
217-244-0072