California State University
Rosta Farzan

Although the ability to use Unix is a necessary prerequisite, many computer science (CS) students have little or no prior experience in Unix before they enter college. I am developing an online Unix tutorial in the form of a meta-tutorial that will accomplish two goals: 1) help students understand Unix concepts and how to make use of online resources, and 2) remove some of the barriers that impede equity in computer science programs.

California State University
Lalitha Balasubramanian, Ravikarthik Sivaswamy, Kalagee Shah
Generating Efficient Test Cases from UML Use Cases

Software testing is one of the most difficult phases of the software development life cycle. Test case selection is a challenging area in the testing field. The aim of the paper is to develop an effective approach in generating efficient test cases from UML use cases. Our approach will be validated through two case studies, which will show how the input domain is optimized, and software requirements are effectively tested.

George Washington University
Arwa Al-Aama
Vigilance in Safety Critical Monitoring Systems

This research defines Safety-Critical Monitoring Systems (SCMSs) as automated systems where critical complex data is monitored for many hours by human operators in order to avoid some hazard. The research identifies vigilance as a major contributing factor to the overall safety of SCMSs. It offers a Vigilance Taxonomy which includes factors that affect human vigilance, their effect, and their references in the literature, if available. The Taxonomy is aimed to serve in identifying new opportunities for researchers and in guiding designers towards more vigilance supportive systems. The research later provides an exemplary prove of how the Taxonomy can be used.

Massachusetts Institute of Technology
Vibhav Rangarajan, Kevin Wilson, and Neal Checka
Source Separation using Audio-Video Sensor Fusion

One goal of a pervasive computing environment is to allow the user to interact in a natural way with the computer. The use of spoken commands is a natural way for humans to interact with many systems. In order to avoid the use of extra equipment for each user, a microphone array can be used to steer to certain locations in the room. The existence of multiple speakers or large interfering signals requires advanced beamforming techniques which attempt to separate the source from the rest of the signal. In this paper, a method based on two such techniques, adaptive and geometric, is developed for source separation in a pervasive computing environment.

Sam Houston State University
Dino Turino Setiawan
Project Erasmus: An Implementation of Online Dynamic Multimedia Tutorial

Project Erasmus is a project to display and to store a variety of multimedia data. The current implementation is to simulate an electronic library, which contains many books, mostly to simplify the hierarchy and the visualization of the problems. The concept of books explains the break down of chapter, page, image, video, and sounds. A remote client can view the electronic book after inquiry, and can store data at the server in an XML format. XML hierarchy attribute explains the functionality of the data. The advantage of project the final result is dynamic, which can be change at any moment from the database.

State University of New York at Buffalo
Philip R. Ventura, Jr.
On the Origins of Programmers: Identifying Predictors of Success for an Objects-First CS1

In 2001 the IEEE/ACM Joint Task Force on Computing Curricula legitimized the teaching of object-oriented programming in the introductory course sequence, termed objects-first. This research looks at an instance of the objects-first approach. It details how an objects-first approach might look. The research examines, through scientific analysis, various predictors of success for an objects-first approach and compares results with more traditional imperative-first approaches.

University of California, Riverside
Jing Li
Efficient Rule-Based Haplotyping Algorithms for Pedigree Data

We study haplotype reconstruction under the Mendelian law of inheritance and the minimum recombination principle on pedigree data. We prove that the problem of finding a minimum-recombinant haplotype configuration (MRHC) is in general NP-hard. This is the first complexity result concerning the problem to our knowledge. An iterative algorithm based on blocks of consecutive resolved marker loci (called block-extension) is proposed. It is very efficient and can be used for large pedigrees with a large number of markers, especially for those data sets requiring few recombinants (or recombination events). A polynomial-time exact algorithm for haplotype reconstruction without recombinants is also presented. This algorithm first identifies all the constraints based on the Mendelian law and the zero recombinant assumption, and then enumerates all possible consistent haplotype configurations. We have tested the block-extension algorithm on simulated data generated on three pedigree structures. The results show that the algorithm performs very well on both multi-allelic and biallelic data, especially when the number of recombinants is small.

Carleton College
Sebastian Celis
Weka-Parallel: Machine Learning in Parallel

We present Weka-Parallel, which is a modification to Weka, a popular machine learning software package. Weka-Parallel expands upon the original program by allowing one to perform n-fold cross-validation in parallel. This added parallelism causes Weka-Parallel to demonstrate a significant speed increase over Weka. Weka-Parallel is designed for the researcher who needs to do intense cross-validation calculations and wishes to transparently and simply harness the power of multiple computers. All the details of cross-validation, result aggregation, and multiprocessor communication are completely handled by Weka-Parallel.

The College of New Jersey
Jessie Burger
Experimental Determination of Size of Data Sets for Support Vector Machine Classification

There has been much mathematical discussion of the necessary size of data sets for support vector machines. There is little published experimentation comparing the classification accuracy of support vector machines created by various sized data sets. This paper discusses such an experiment to determine the number of vectors and the number of attributes needed to create an accurate classifier. The most important aspect of a training set is the balance between positive and negative vectors. A fairly accurate support vector machine can be created using both few vectors and few attributes if the training set is balanced.

Duquesne University
Michael Piatek
An Analysis of Flow Identification in QoS Systems

The increasing reliance on network services at large organizations has prompted administrators to seek means of ensuring resource availability for critical applications. Although several commercial offerings are presently effective, they cannot guarantee performance in the long term because of a serious vulnerability in their underlying architecture. We present a scalable, computationally practical method of masking the protocol structure of any network data through a modified encryption technique. This procedure renders traffic invisible to current real time classification techniques.

Furman University
Michael D. Elder
Distributed STrategy-Oriented Recursive Messaging System (Distributed STORMS): A Platform for Rapid Java-based Distributed Application Development on Beowulf-class Supercomputers

The design and implementation of cluster-based distributed applications can be daunting for the uninitiated. Distributed STORMS provides a reusable platform for rapidly developing cluster-based distributed applications using Java and the Extensible Modeling Language (XML). It provides for the development of pluggable, reusable Java components which can be molded into XML-based instruction lists. Once formed, these components are organized into recursive messaging hierarchies for improved performance. The result is a system which allows users familiar with Java, but new to distributed development, to experiment with minimal development time and effort.

Furman University
Jon Wyrick
ENTITY: A Component Software Architecture For Developing Scientific Simulations

I have developed a new architecture that I call ENTITY for writing applications. It makes use of Java reflection and the JavaBean API in order to wrap around program components and make them a live part of the application framework. Using ENTITY I have written an application called JavaWave that allows scientists to interactively "wire up" simulations of atmospheric models and then visualize the results. JavaWave will soon be released to the NASA-TIMED satellite team for analysis of data from the TIMED satellite mission.

Hope College
Agata Bugaj, Lilyana Mihalkova, and Donald Porter
Using Java to Teach Networking Concepts With a Programmable Network Sniffer

In a networking class, it can be difficult to master the subject matter without an opportunity for "hands-on" experimentation, such as examining a network and interpreting the applications and protocols running across it. NetSpy is a platform for packet sniffing whose unique features make it excellent for classroom use. NetSpy captures packet data and organizes it in an intuitive object-oriented manner. It abstracts the details and allows users to define packet filters as well as their own modules for network data analysis. NetSpy's GUI has been specifically designed to allow NetSpy to be used on a hand-held device.

Hope College
Alexander A. Sherstov
Streamlining Distributed Application Development

DisViz is a general-purpose lightweight execution environment for distributed applications that provides easy-to-use facilities for distributed scheduling and execution control. DisViz is equipped with an extensive utility library supporting peer-to-peer connections, unicast/broadcast messaging, and distributed synchronization. DisViz supports applications in such diverse areas as hardware simulation and testing, distributed AI simulations, and visualization of distributed computation. Among DisViz's major assets are versatile functionality, extensive failure-handling facilities, high performance (as measured by processing efficiency, bandwidth usage, and responsiveness), ease of use, scalability, and portability.

St. Edward's University
Melissa Y. Zavala
Can Genetic Programming Techniques Improve the Performance of Multicast Routing Algorithms?

Based on concepts from the biological sciences, genetic programming techniques have been used to create more efficient algorithms in many areas of computing. In my research, I applied genetic programming techniques to the problem of creating a multicast routing tree. I compared my modified algorithm against two other well known multicast routing algorithms. Results indicate an increase in speed but a marked decrease in accuracy. Complete details on this project may be found at http://acad.stedwards.edu/~mzavala/4247.

Southwestern University
Ryan Smith
Functional Ham

The problem that I addressed was to implement a generalized Hamming encoder and decoder in a functional programming language. The program is able to encode a 2^n-n-1 bit binary number into a 2^n bit binary codeword using a slight variation on the original Hamming coding scheme; it can also decode any 2^n bit binary word, and correct single errors and detect dual errors. The code works for any n>2.

University of Alabama, Birmingham
Musawir Ali Shah
Learning and Generalizing Foraging Mechanisms Through Bio-inspired Neural Synaptic Darwinism

Adaptive learning is a very important part of artificial intelligence. The ability of an agent to explore and adapt to its surroundings, adjusting itself in the process is a helpful aspect which can lead to automated terrain examination (for example) and such other practical applications. In common practicality, the theory of adaptive learning can be applied to creating self sustained machines which can be set to do a specific task without manual input at each stage of the course. This study combines adaptive learning and the generalization of foraging and survival methods by emulating the biological neural system. In the study, a naive agent is placed on a map (habitat), which consists of white-spaces (spots on the map where the agent can move to), obstacles, and nutrition sources. The agent will develop an understanding of its habitat, and will begin to exhibit intelligent behavior by applying different foraging techniques; moving towards nutrition sources when the need arises, or exploring un-explored parts of the map when its energy is at an optimal level.

University of Nebraska, Lincoln
Nathan Wells
Development of the Self-Calibrating Palmer Drought Severity Index

The Palmer Drought Severity Index (PDSI) is one of the standard methods used in quantifying drought conditions. It has several inherent problems, all caused by the use of empirical constants in the algorithm for computing the index, that result in the inability to compare the index spatially. The Self-Calibrating PDSI is a refinement of the PDSI that does not use any empirical constants. In their place, values are dynamically calculated to calibrate the behavior of the index based on location specific historical climate data. The result is that the Self-Calibrated PDSI behaves consistently and predictably at any location.

University of Northern Iowa
Ryan Dixon
Fontango: A Tool for Creating Personal Fonts from Handwriting

Fontango is a font creation tool designed for Mac OS X and based on Apple's Cocoa frameworks. The goal of this project is to produce an application capable of converting human handwriting into a computer-based font; a program that is able to emulate the random nature of human writing. Current research includes investigations into curve approximation, data input, data extrapolation, graph optimization, and human-computer interaction.

University of Wales, Aberystwyth
Wayne Ellis
Capturing Collaborative Designs to Assist the Pedagogical Process

Growing research shows that the pedagogical process of Object Oriented Programming is flawed, with an average failure rate of as much as 30% of freshman unable to grasp OOAD concepts. Many Universities use Integrated Development Environments (IDEs) to assist novice programmers, yet studies show that IDEs are possibly making things worse. My research involves building an IDE(VorteX) to assist freshman students in their understanding of the concepts of OOAD and at the same time captures every action performed by the student. In this way we can both study the evolution of a student's design and ultimately use case-based reasoning to allow VorteX to provide freshman with automated assistance.

University of Wales, Aberystwyth
Jonathan Holloway
Development of a Managed Learning Environment by Students for Students

TWEEK is a managed learning environment that is both platform and system independent and designed to be a completely modular learning environment that is secure, robust and above all inter-operatable with other MLE systems. Developed by students for students in a university environment it is designed to be a completely innovative environment that closely models the e-learning experience. It incorporates the advantages of the current learning process whilst eliminating many of the disadvantages such as accessibility and distance learning.

Peter Sanderson, Otterbein University
Anne Gates Applin, University of Southern Mississippi
Ian Utting, University of Kent at Canterbury
Joe Hollingsworth, Indiana University, Southeast
Dennis Bouvier, St. Louis University
Aaron Gordon, Metropolitan State College of Denver
Dan Joyce, Villanova University
Mark Lewin, Microsoft Research
David J. Barnes, University of Kent at Canterbury
Ellen Walker, Hiram College
Lynda Thomas, University of Wales, Aberystwyth
Doug Baldwin, State University of New York, Geneseo
Larry Merkle, Rose-Hulman University
Judy Williams, William Penn University
Chris Milner, University of Virginia
Russell May, Morehead State University
Martha McCormick, Jacksonville State University
Cate Sheller, Kirkwood Community College
Jim Arman, Columbus School for Girls
Jody Paul, Metropolitan State College of Denver
Dorothy Deremer, Montclair State University
Clark Archer, Mount Union College
Jack Rehder, University of Waterloo
Jasmin Chadha, University of South Carolina, Spartanburg
Andrew Staugaard, College of the Ozarks
T. LaVonne Manning, University of D.C.
Robert Moutante, Bloomsburg University
Gail T. Finley, University of D.C.
Steven Cooper, Saint Joseph's University
Mark Ratcliffe, University of Wales, Aberystwyth
Gloria Melara, California State University, Northridge
Mark Bailey, Hamilton College
Prabhaker Matetii, Wright State University