GHC - 2004

POSTER PRESENTATIONS

Using Autonomous Robots to Integrate Middle School Academic Standards, Technology Standards, and 21st Century Workplace Skills

Barbara Burckart - Teacher, Nova Southeastern University

The amount of knowledge that students are required to master continues to grow. Teachers struggle to find age appropriate ways to integrate state academic standards, technology standards, and 21st century workplace skills. Improved problem solving skills are deemed important to these standards as well as in the workplace. This dissertation utilized a robot design project to examine the changes in problem solving skills of seventh grade science students. In the process, 21st century workplace skills and science, math and technology standards were identified and integrated into the project.

Compiler-Assisted Efficient Checkpointing

Alison Norman - Graduate Research Assistant, The University of Texas at Austin
Sung Eun Choi - Research Scientist, Los Alamos National Laboratory
Calvin Lin - Associate Professor, The University of Texas at Austin

Supercomputer clusters can consist of thousands of processors. If even one processor fails, the application must restart from the last checkpoint. Currently, checkpoints are inserted manually by application programmers, so all processes checkpoint synchronously. However, as clusters and applications grow, these synchronous checkpoints will suffer increased contention in both the network and the file system. We present a compiler-based technique that significantly reduces contention by placing individual processes' checkpoints at different points in time. By relying on the compiler, our solution reduces the fault-tolerance burden from the programmer.

The Career Development of Women in Executive Level Positions in Information Technology

Rose Mary Wentling - Senior Research Scientist, National Center for Supercomputing Applications (NCSA), University of Illinois
Steven Thomas - Research Associate, University of Illinois

The information Technology (IT) field is male dominated and particularly at the executive level position. In addition, it is affected by the shrinking number of women pursuing academic study in computer science and engineering, both at the undergraduate and advanced-degree levels. The recent sharp decline in the number of women pursuing undergraduate degrees in computer-related fields and the attrition of women in advance-degree programs affect the number of women at levels higher in the pipeline in IT. This study examined the career development and aspirations of women in executive level positions in IT.

Graph-based Clustering of Biological Data

Aynur Dayanik - Ph.D. candidate, Rutgers University
Craig Nevill-Manning - Director, Google

Biological databases contain a wide variety of data types with rich relational structure. We describe a novel method for clustering biological data by exploiting this relational structure. First, we construct a network of sequence-structure-literature with pairwise relationships. Then we apply graph partitioning techniques to infer clusters of related articles, sequences and structures. We found that the resulting clusters exhibit strong topicality.

Assistive Technologies for Dialysis Patients

Katie Siek - Research Assistant, Indiana University
Kay Connelly - Assistant Professor, Indiana University

Dialysis patients can only consume 1 liter of fluid and two grams of sodium each day. Currently, patients try to remember or write down in a food diary their fluid and sodium consumption. However, these techniques are insufficient because 80% of patients are unable to restrict their fluid intake. If patients miscalculate their fluid intake they run the risk of hypertension, pulmonary edema, and death. Our research focuses on creating a personal digital assistant application to assist dialysis patients accurately monitor their fluid and sodium intake.

Congruential Sieves on a Reconfigurable Computer

Heather Wake - Research Assistant, University of South Carolina
Duncan Buell - Professor and Chair, University of South Carolina

Computationally dense applications take a long time to run on conventional computers. Such applications are better suited to run on custom processors whose hardware has been specifically designed for the application. As an alternative to dedicating hardware forever to a particular computation, reconfigurable hardware can be designed for the application and then later redesigned to solve another problem. A congruential sieve is the ideal example of an application which is better suited to run on hardware than on software. We describe here an implementation of a sieve on the Star Bridge Systems HC 36m reconfigurable computer.

Generational Cache Management of Superblocks in Dynamic Optimization Systems

Kim Hazelwood - Postdoctoral Researcher, Harvard University

Dynamic software systems store optimized/translated code in a software-managed code cache in order to maximize reuse and amortize overhead. Just as with hardware-based caches, code caches must include a cache management policy to determine when and how to evict elements from the cache. Yet, unlike hardware caches, code caches store elements that are not fixed in size, may contain links to other elements, and have a very high overhead for premature eviction. We propose and evaluate a new cache management algorithm based on generational code caches.

DistDLB: A Dynamic Load Balancing Library for Large-scale Applications on Distributed Systems

Zhiling Lan - Assistant Professor, Illinois Institute of Technology
Yawei Li - Graduate student, Illinois Institute of Technology

We present DistDLB, a dynamic load balancing library for large-scale adaptive applications on distributed systems. DistDLB performs load balancing through an innovative combination of a hierarchical load balancing approach with real-time performance modeling. That is, load balancing is performed in a two-phase manner and the balancing decision is chosen adaptively at runtime based on the observation of system performance and application state. The resulting DistDLB library can be easily used to simplify and optimize the development of distributed applications.

A Framework for Greedy Methods to Construct Interaction Tests

Renee Turban - PhD Student, Arizona State University

Several greedy algorithms exist for constructing interaction test suites. These algorithms are measured on three major criteria: accuracy, consistency, and efficiency. Further consideration may be given to their extensibility and practical application to testing concerns. This research makes two contributions. A framework for greedy methods to construct interaction tests is defined and experimental results of the statistical contributions of the features within this framework is provided.

Role Models: Making a Difference in a Girl's Future and in the Future of Technology

Linda Kekelis - Project Director, Chabot Space & Science Center
Jeri Countryman - Project Coordinator, Chabot Space & Science Center
Etta Heber - Programs Director, Chabot Space & Science Center

Imagine that the design for a fuel-efficient car is in the mind of a girl? The discovery might never be made since many girls say “no thanks” to computers. Through personal stories and engaging activities, role models can spark girls' interest in technology. In our poster, we will present resources to train role models from Techbridge, an afterschool technology program for girls hosted by Chabot Space & Science Center in Oakland and the California School for the Blind. While there are many programs that introduce role models to youth, few report on the kinds of support required for their success.

Spotlight: Detecting and Mapping Unresponsive Paths in IP Networks

Margaret Richey - PhD Candidate, University of Wisconsin Madison
Hetal Pathak - Graduate student, University of Wisconsin Madison

For better understanding, management and measurement of networks, knowledge of the topology is important. The objective of our tool, ‘Spotlight' is to shine a light on the dark regions in the Internet by eliciting response from the unresponsive routers using TCP, UDP, ICMP and MTU discovery. We describe the design of the tool with respect to placement of the monitors and IPs, network MTU and router access lists. The second phase is to infer the topology of the network based on IP identifiers. Presently, the tool has been able to elicit response from more than half of the unresponsive IPs.

Signal-Specialized Parameterization for Piecewise Linear Reconstruction

Geetika Tewari - Graduate student, Harvard University

We propose a metric for surface parameterization specialized to its signal that can be used to create more efficient, high-quality texture maps. Our metric predicts the signal approximation error - the difference between the original surface signal and its reconstruction from the sampled texture. Unlike previous methods, our metric assumes piecewise-linear reconstruction, and thus makes a good approximation to bilinear reconstruction employed in graphics hardware. We achieve significant savings in texture area for a desired signal accuracy compared to the signal-specialized parameterization metric proposed by Sander et al. in the 2002 Eurographics Workshop on Rendering.

Preparing Students for a Global Economy

Sharon Perkins Hall - Associate Professor, University of Houston - Clear Lake

Since outsourcing of information services has begun to impact the ability for undergraduate computer science students to find employment, universities are faced with the dilemma of preparing students to compete in a global economy. Partnerships with industries such as manufacturing that have already met this challenge and have been successful in competing and cooperating in global markets can help to close the competency gaps for computer science students. A first step at understanding how to better prepare students for a global economy in the computing fields is to address globalization as part of a senior project class or seminar.

Recognizing Spoken Directives Using a Fully-Statistical Denotational Language Model

Shana Watters - Graduate Student, University of Minnesota
William Schuler - Assistant Professor, University of Minnesota

Computers are such an integral part of our daily lives that they can be found nearly everywhere. Most users interact with computers via a keyboard and mouse but what they really want is a more user friendly and hands free interface such as a spoken language interface. Our prototype system takes a test subject's spoken directive to "pick up" an object and attempts to correctly predict what entity the test subject wants by using a denotational language model that conditions probability estimates of candidate analyses on information such as an entity's color or spatial relationships with other entities.

Minimal Models for Dual Heuristic Programming

Sangsuree Vasupongayya - Researcher, Portland State University
Roberto Santiago - Laboratory Manager, Portland State University
Thaddeus Shannon - Senior Researcher, Portland State University
George Lendaris - Professor, Portland State University

Dual Heuristic Programming (DHP) technique is a soft-computing method that implements approximate dynamic programming which is a general approach for designing an optimal controller for a nonlinear dynamical system given an arbitrary goal or target. We show that only minimal models are required for accomplishing successful DHP. A minimal model is a minimal set of derivatives that are specified to be non-random together with a specification of how non-random they need to be. If an “appropriate” subset of derivatives can be selected for a given dynamical plant, a full system identification process may not be needed.

Constraint-Based Layout Using a “Boxes and Glue” Model

Sangsuree Vasupongayya - Graduate Student, Portland State University
Bart Massey - Assistant Professor, Portland State University

A two-dimensional layout constraint optimization problem is an arrangement of two-dimensional layout objects in a layout space that respects layout constraints and minimizes some cost function. In this work, we cast the original problem as a constraint satisfaction problem to be solved using combinatorial search techniques such as local and system search. The constraint model is based on containment of boxes in a rectangular frame, with glue objects constraining the relationship between boxes. Our model generalizes the boxes and glue model of TEX as well as spring-mass models, illustrating its intuitive appeal.

Requirements-Based Test Case Prioritization

Hema Srikanth - Doctoral Student, North Carolina State University
Laurie Williams - Assistant Professor, North Carolina State University

Test case prioritization techniques have been shown to improve regression-testing activities by increasing the rate of fault detection, thus allowing testers to fix faults earlier. The current techniques, mostly code coverage-based, treat all faults equally. We build upon this work to develop a prioritization scheme with two main goals: identifying the severe faults earlier and minimizing the cost of test case prioritization.

Service Balancing of Wireless Networks

Susan Lincke-Salecker - Asst. Prof., University of Wisconsin Parkside

Engineering traffic across different wireless networks via handovers can increase capacity, flexibility, modularity, and survivability. This work investigates Adaptive Placement, which rearranges traffic, via handovers, when specific networks saturate. Since emerging networks serve circuit-oriented (e.g., speech) and queued packet services of varied data rates, networks must balance queuing delays against blocking rates. Traditionally ‘Load Balancing' is used to equalize processing loads across processing systems. A more appropriate goal for wireless networks is ‘Service Balancing', whose goal is to ensure that service-specific blocking and delay goals are achieved, regardless of any networks' loading level.

Proposing CAPS as a Link in the Bridge across the Divide

Cynthia Lester - Graduate Teaching Assistant, The University of Alabama
Marcus Brown - Associate Professor of Computer Science, The University of Alabama

The decline of women in technology disciplines has been given many different labels in current literature such as "the pipeline shrinkage," and the "gender divide." Research in the area of gender differences in computer-related use and behavior has grown significantly. While there are numerous suggested reasons for the decline of women in technology-related disciplines, the perceptions that women hold about their computing capabilities continue to be of interest. The research seeks to explore how a Computer Assisted Peer-modeling Software application (CAPS) can be used to influence females' level of computer self-efficacy and performance and hence affect their perceptions on computing.

Computer vision-based surveillance of traffic intersections

Harini Veeraraghavan - Graduate Student, University of Minnesota

With increasing concerns for security, an automated surveillance system, capable of detecting interesting events by monitoring scenes would be very useful tool. While a vision-based solution is very good for making scene inferences, it faces several challenges when deployed in an uncontrolled, outdoor environment such as traffic intersections. Ambiguities introduced by occlusions, lighting variations, perspective distortions, etc, affect target registration and thereby, scene interpretation. In this work, we combine multiple cues such as motion-based segmentation, color, and a geometric appearance model of the targets to improve registration and tracking. Results of successful tracking and incident detection are presented.

VoiceNotes: A Discourse-Theoretic Analysis of Programmer Voice Annotations

Jinger Zhao - Graduate Student, University of California , Berkeley

Poor program documentation makes code hard to understand and maintain. Enabling programmers to annotate their programs by voice may improve program documentation. To design a voice annotation tool, however, programmers' voice commenting behavior must be understood. A discourse theoretic method of analyzing programmer voice comments is developed and applied to an empirical study of programming recorded using VoiceNotes, an audio capture and playback plugin for Eclipse. The results of the analysis and their implications for the design a voice annotation tool are discussed.

Computer Security Features on Windows Platform

Wai Yan Elsa Tai - Student Researcher, University of Texas at El Paso

The Windows XP operating system has many security features and policies. These are termed the Windows security space in this research. An analysis and evaluation of the Windows security space is conducted. Based on this analysis, it is concluded that regular operators require additional skills, knowledge and time to deal with the complexity of the security space. Recommendations are given. The Unified Management Model is proposed as a new security model to address some of the identified problems, including costs. This research aims to improve the ease of hardening computer security space, thereby raising the security level in cyberspace.

Bioinformatics for Genome Analysis: multiple sequence alignment with fastDNAml

Gloria Rendon - Sr. Research programmer, Dept. of Microbiology, University of Illinois at Urbana-Champaign
Gary Olsen - Professor, Dept. of Microbiology, University of Illinois at Urbana-Champaign

The number of publicly available complete genome sequences will continue to increase for decades to come. Handling this volume of data requires automation of many tasks. We propose to provide improved tools for Genome Analysis by more tightly coupling presumed evolutionary relationships of proteins (phylogenetic trees) with the generation of multiple sequence alignments. fastDNAml is a program developed at the University of Illinois at UC that performs phylogenetic tree inference using a maximum likelihood tree inference method with features that are useful for large tree. The poster presented at GHC '04 will display the major highlights of the fastDNAml program.