Poster Session - Thursday, 7:00pm to 10:00pm

Poster abstracts are alphabetized by first author listed.

Hamiltonian Systems
Stephen Bond and Michael T. Turnley, Siebel Center for Computer Science, turnley2@uiuc.edu

Many systems in nature are governed by conservative processes, including gravitational N-body dynamics, rigid-body motion and classical molecular dynamics. The time scales of these systems range from the order of milliseconds to millions of years. To simulate these dynamical systems efficiently, the numerical methods used to approximate solutions to these systems must be designed to respect the qualitative properties (invariants) of the underlying systems. For conservative Hamiltonian systems, geometric integrators are very effective since they preserve many of the underlying invariants of the motion. The research presented in this poster explores the properties of geometric (e.g. symplectic) integrators for resolving long time trajectories of Hamiltonian systems. This study makes direct comparisons between known geometric integrators to understand the relative efficiency of various methods appearing in the literature.

Visualization of Human Tissue Model for Space Radiation Dose Assessment
Kevin Calvin, Prairie View A&M, kevincalvin@hotmail.com

The science of the space radiation environment and its transport through the human biological systems is more complex than the most recent advancements of proton and heavy ion radiation treatment plans. The aspect of space radiation transport not only includes the primary high energetic particles from the free galactic space, but also the secondary particles generated in the shielding shell of the space vehicle. Though little, human tissue offers several gradations of self shielding for the radiation environment as a function of the tissue density and depth. In this paper we present the human tissue model with more realistic representation of tissue density variations based on the visible human body data. Our first attempt is with the visible human body head data set. We present our reconstruction of the 3D shield model for radiation transport and dose calculations based on the CT, MRI and anatomical pictures of the visible human body cross-sectional data.

This research is sponsored by the NASA grant through the NASA-Center for Applied Radiation Research at PVAMU.

Variations of Radiation Environment during the Current Solar Cycle using CRIS Data
Tausha Calvin, Prairie View A&M University, tausha_calvin@yahoo.com

The Advanced Composition Explorer (ACE) of NASA, launched in 1997, has been in orbit at the Sun-Earth libration point (L1) ~1.5 million km sunward of Earth. Several high-resolution spectrometers onboard the ACE measure the elemental, isotopic and ionic charge state composition of nuclei from H to Ni (1≤Z≤28) from solar wind energies (~1 keV/nuc) to galactic cosmic ray energies (~500 MeV/nuc). Data from the Cosmic Ray Isotope Spectrometer (CRIS) instrument are being analyzed to understand the variations in the particle spectrum during the current solar cycle phase. Understanding the variation of the background galactic cosmic rays particle flux over the time period between solar maximum and solar minimum in the near earth orbit is an important step in assessing the space radiation risk for deep space human exploration missions. We present the radiation particle flux variation in the current solar cycle from the ACE/CRIS measurements compared with model calculated predictions.

This research is supported in part by a NASA grant through the NASA-Center for Applied Radiation Research at PVAMU.

Service-Oriented Architecture for Sensor Networks
Brian Carter, University of Louisville, brian.carter@louisville.edu

Sensor networks are delivering near-real-time information to scientists worldwide. Extracting this information to gain knowledge and understanding is one of the greatest challenges faced today. From national defense, medical applications, to the environment, the data delivered from the sensor networks are unstructured, using their own format and protocols.

Our research presents a service-oriented architecture to comply with emerging Grid and Web service standards. Aggregating the information from sensor networks using existing standards will increase the quantity and quality of the data available. This will allow for emerging devices with very different sensing and computation capabilities to interact and exchange information in networks that have far greater breadth of application than today’s traditional Grids.

We will present the research being conducted showing the integration of Web standards. The poster will reflect current research from SensorNet, Sensor Model Language (SensorML), Federal Enterprise Architecture (FEA), World Wide Web Consortium (W3C), and sensor architecture research conducted at our lab facilities. Our research will present our contribution to unifying several standard groups for the next generation service-oriented architecture.

Dispelling Computer Science Gender Myths: Perceived Motivation, Persistence, and Confidence Levels Among Undergraduate Men and Women
Belinda Chang, Catherine Fan and Carol Frieze, Tufts University, cfan@andrew.cmu.edu

This study assesses some of the changes in the culture of computing at Carnegie Mellon University since 1999, specifically addressing gender similarities and differences as they relate to perceived motivation, persistence and confidence levels among undergraduate computer science students. Analysis and results are based on data from three sources: interview transcripts from the graduating class of 2004, surveys of the current student body, and surveys of students who transferred out of computer science. Our findings revealed some strong gender similarities among men and women, challenging the notion of a strong gender divide in how undergraduates relate to computer science. New understanding and insights gained from this ongoing investigation will be used to produce recommendations to enhance the overall success of undergraduate students. Such recommendations will target women students in particular, as the issue of attracting and retaining women in computer science continues to be a national concern.

Online and Offline Power Estimation Techniques for Intel Xscale Technology Cores
Gilberto Contreras and Margaret Martonosi, Princeton University, gcontrer@princeton.edu

Recent years have seen a proliferation of embedded devices in many aspects of life, from cell phones to automated controllers. These devices are typically characterized by their low power consumption, small profile and ever-increasing performance demands. Under such constraints, understanding power consumption of running software is of extreme importance since it can help us perform early power and performance design optimizations. This work presents two independent approaches for estimating power consumption of an embedded processor. The first approach is an offline technique that uses a functional simulator and a set of parameterized mathematical power models to estimate power consumption of the processor. The second approach is a runtime linear power estimation technique that uses hardware performance counters to estimate CPU and memory power consumption. These methods prove to be very accurate, presenting an average error of only 4% each with respect to physical power measurements. Overall, the tools offer significant promise in helping hardware and software engineers to predict and optimize system power consumption.

Lets Vote: A Multimodal Electronic Voting System
E. Vincent Cross II and Juan Gilbert, Auburn University, crossev@auburn.edu

After the debacle of the 2000 U.S. presidential election, it became abundantly clear that America’s archaic voting system was in need of a major overhaul. Consequently, the federal government has allocated funds through the Help America Vote Act (HAVA) for the purchase of modern voting equipment. Specifically, electronic voting (E-Voting) machines are the type of modern voting equipment which are under consideration for use in a number of states. Incidentally, E-Voting machines are currently being used in a select few states; however this has not been without controversy. This poster describes an electronic voting system that has been developed to address many of the aforementioned issues. This system is multimodal, meaning that it accommodates speech and touch interactions. Some of the key features are the ability to print bar codes and textual votes, network and stand-alone functionality, increased security and support for three distinct methods of recounts.

Laptop Robotics: Low-Cost Access to Research Robots
Zachary Dodds, Harvey Mudd College, dodds@cs.hmc.edu

Robotics has been demonstrated to draw students from all walks of life toward computing: increasing numbers of high schools and undergraduate CS departments have explored robotic kits to introduce ideas fundamental to computation, programming and spatial reasoning.

Unfortunately, the initial excitement of Lego robot platforms fades too quickly to frustration. In about three weeks, students realize that (1) robust mechanical design with Lego's is difficult and is a Lego-specific skill, i.e., does not transfer outside the endeavor of Lego robot-building and (2) the computational capability of Lego robots limits the tasks they can handle substantially.

We have turned to very low-cost (<$300) laptop-driven platforms to remedy these problems. The result has been a scalable pedagogical approach to robots: inexpensive and simple enough for novices, but powerful enough that students will not outgrow it.

This poster will present our laptop robot and the student projects that it has enabled.

Because these robots leverage a department's or school's existing laptop computer, they provide a substantially more sophisticated experience than Lego-based kits. In addition, there is no special-purpose interface to learn: students can work in Windows, MacOS (FreeBSD), or other Linux/Unix systems using any language and development environment.

Efficient Finite Field Arithmetic for Field Programmable Gate Arrays
Edgar Ferrer, Dorothy Bollman and Oscar Moreno, University of Puerto Rico, eferrer@cs.uprm.edu

Finite field arithmetic has a wide range of applications in various fields, including cryptography and error-correcting codes. This project is motivated by a novel method modeling genetic networks by means of finite fields. In order to deal with very large genetic networks, it is essential to develop capacity for performing very fast and efficient arithmetic over finite fields. For this purpose we are developing efficient algorithms that will be hardwired into FPGAs. One very efficient method for arithmetic on GF(2m) involves the use of Zech logarithm tables. This method is very efficient for small m. For large composite m, say m = rs , GF(2m) is an extension of GF(2r) and we can make use of the fact that GF(2m) is isomorphic to GF(2r)s.. In this project we make use of the Zech log table method for arithmetic over the ground field GF GF(2r) and FPGAs for arithmetic over the extension field.

Efficient Distance Field Computations using Cluster Trees
Sarah Frisken and Elena Jakubiak, Tufts University and Ron Perry, Mitsubishi Electric Research Laboratories, jakubiak@cs.tufts.edu

Rendering fonts on small display devices such as cell phones poses new challenges. To fit sufficient information on tiny display devices, small font sizes must be used and fonts must be rendered with high clarity so that text is easily readable. Adaptively Sampled Distance Fields (ADFs) provide a new way to render such fonts. Initially, this method was slower than desired. To render a glyph (a letter, number or symbol) using an ADF requires repeated costly distance computations. Many of the distance computations, however, are ultimately unnecessary. Cluster Trees use hierarchical clustering to gain an idea of the layout of a glyph and use this to prune unnecessary distance computations. The use of Cluster Trees for ADF font-rendering resulted in a 10x speed-up. This new font rendering technology will be included in the next release of Macromedia Flash Player.

Text Entry for Automobiles
Iván E. González, Jake Wobbrock and Brad A. Myers, Carnegie Mellon University, ieg@andrew.cmu.edu

The need for text entry in an automobile setting has become a new challenge as more devices that require user input are integrated with the driving experience. Some of the solutions that have been used in the past include on-screen keyboards and gestural methods, both of which require the user’s visual attention. The proposed solution is EdgeWrite, a unistroke text entry method that has been previously implemented for a stylus on a touchscreen, a finger on a touch pad, a joystick on a game controller or wheelchair, a mouse, trackball or TrackPoint connected to a regular computer. Gesture recognition in EdgeWrite is accomplished through the sequence of corners that are hit in a square region. The need of only four binary sensors allow high tactility and can alleviate the need for visual attention making it an efficient solution for text entry in automobiles. Two prototype steering wheels were developed that integrate EdgeWrite, one with a small touch pad, and another one with four buttons. Steering EdgeWrite will be compared using a driving simulator to previous studies that measured effects of talking on a cell phone while driving.

A Molecular Dynamics Study of the Role that Surface-Active Phospholipids play in the Lubrication of Synovial Joints: Implications for the Treatment and Prevention of Osteoarthritis
Stacie LeSure Gregory, North Carolina State University, slesure@hotmail.com

The superb lubrication and low friction associated with synovial joints are partly due to the presence of synovial fluid. The physiological function of the joints is severely comprised if this fluid is deteriorated, possibly leading to osteoarthritis. Surface active phospholipids (SAPLs) are the primary components lending synovial fluid its lubricating properties. This research effort is using molecular dynamics simulation to understand the actual role that SAPLs play in joint lubrication. There are two major issues being investigated. The first study will determine how these double chain fatty acid molecules behave when two opposing surfaces containing SAPLs are brought into close proximity. Secondly, single chain and triple chain molecules will be modeled in an effort to understand why nature chose the double chain fatty acid molecules to act as boundary lubricants. Single or triple chain molecules may prove to have beneficial pharmaceutical applications.

Speech Usability Metric: Evaluating Spoken Language Systems
Priyanka Gupta, Auburn University, guptapr@auburn.edu

This poster presents a speech usability metric to evaluate spoken language systems. This evaluating method tests spoken language systems on three levels, namely user satisfaction, accuracy and task completion time. It allows the designer to place emphasis on the specific metrics that according to him/her are most important for designing a highly usable system. Currently, PARADISE (PARAdigm for Dialogue System Evaluation) is the most-cited method for evaluating spoken dialogue systems. It is based entirely on maximizing user satisfaction. However, there are some instances where user satisfaction is not paramount. In these cases and those where user satisfaction is paramount, the speech usability metric can be used. An example of such a system could be a spoken language system where accuracy or task completion time is paramount, e.g. military applications. The speech usability metric addresses all the same issues as the PARADISE Framework, but it is less complex and does not require a large set of users to test the system. The primary focus of this research was to validate this usability metric as an alternative tool for evaluating spoken language systems. The results of our usability study will be reported.

How Do We Get the Clinical Side of Healthcare On Par with the Administrative Side of Healthcare with Respect To Information Technology Alignment?
Karen Hare, Brian Whitworth and Albert Jekelis, New Jersey Institute Of Technology, kxh1868@njit.edu

Healthcare information systems have historically been administrative systems, with responsibility for financial viability of the healthcare organization. Currently there exists a distinction in the level of automation that has taken place on the administrative verses the clinical side of healthcare. Information technology systems that are utilized for clinical healthcare have unique requirements as well as legislative mandates for patient privacy, confidentiality and security. Coupled with these are the complex nature of the data and the way that they need to be communicated.

A preliminary study has been conducted that suggests that there are non-monetary factors that contribute to the gaps in adoption of information technology on the clinical side of healthcare. Inhibitors were identified that precluded clinicians from embracing the technology that could help them implement the technology solutions. These findings suggest that functionality, user-friendliness and system integration, as well as faster system response time, are major concerns retarding implementation of clinical information technology solutions. This continuing research will seek to identify information technology solutions for clinical healthcare practitioners that address agility and flexibility, while adhering to the constructs that are necessary for secure and confidential data.

Influences of Basket and Stellate Cells on Cerebellar Purkinje Cells
William Holtkamp and William Stegall, University of Houston Downtown, zaecron@gmail.com

The cerebellum is vital to motor actions in everyday activities. In the cerebellum, the Purkinje cells are the only output (consisting of voltage spikes which are also called action potentials) which responds to the rest of the brain. There are two types of cells, basket and stellate, which inhibit the Purkinje cell. Not many studies have focused on these cells, and their functions remain unknown. What do these cells do? Why are they in the body? This study tries to explain these questions, utilizing biophysically detailed computational models developed using NEURON 1. New models were created of basket and stellate cells, using known physiological parameters, and these were incorporated with preexisting models of Purkinje cells. The simulations reveal that basket and stellate cells seem to act as a fuse in order to not allow the Purkinje cells to receive mass amounts of excitation. Our tests prove this theory to be true and in fact prove the interneuron cells (basket and stellate) to have great effects on Purkinje cells to firing rates and excitability.

1 NEURON is a tested program. Information can be found at http://neuron.duke.edu/ and also at http://www.neuron.yale.edu/neuron/.

Towards Coral Bleaching Early Detection using Data Mining
Jerrold Illis, University of the Virgin Islands, ill8225@uvi.edu

In recent years, coral bleaching has become an epidemic claiming the lives of many of our coral reefs around the world. There are various environmental and atmospheric factors that can cause this phenomenon. The most prevailing theory is that mass coral bleaching has occurred in association with episodes of elevated sea temperatures over the past 20 years. When the thermal tolerance of corals is exceeded this leads to the loss of the essential zooxanthellae that reside within the corals. These organisms have a symbiotic relationship with corals and perish without the zooxanthellae and the nutrients they provide. This research considers using K-means clustering algorithms on coral reef data for effective coral bleaching data analysis. Data will be partitioned into groups and clustered based on the selected atmospheric categories (e.g. Air Temp, Sea Temperature, Salinity, Wind Speed, Fluorometry, etc.). The results suggest that clustering algorithms can partition coral reef data into groups based on the similarity between their expression profiles. In this way, trends and relationships can be observed aiding in predicting future coral bleaching events. Observations on sea temperature playing a major role in coral bleaching will be assessed along with the role of accompanying factors.

Improving Object Inlining for High-Performance Java Scientific Applications
Mackale Joyner, Rice University, mjoyner@rice.edu

Java is a popular programming language that enables many developers to achieve high productivity. Previous work in Java improved runtime performance by using object inlining. This research extends prior object inlining work by both analyzing the code and performing optimizations to further improve application runtime performance. Two impediments to object inlining and to increased runtime performance are object and array aliasing and binary method invocations. This research implements object and array alias strategies to address the aliasing problem while utilizing an idea from Telescoping Languages to address the binary method invocation problem. Application runtime gains of up to 20% result from employing these techniques. The improvements made to the compile-time object inlining optimization should increase the scientific community's acceptance of the Java programming language in the development of high-performance scientific applications by decreasing the performance gap between Java and accepted languages such as C and Fortran.

Understanding Interaction Equals Better Performance
Charles Lively, Xingfu Wu and Valerie E. Taylor, Texas A&M University, clively@cs.tamu.edu

The GYRO plasma-physics application has been introduced as a key tool to quickly analyze and solve various physics scenarios. However, previous work has shown that in order to improve the performance of this application, a better understanding of the collision step, including kernel interactions, must be studied. This work analyzes the performance of the GYRO plasma-physics application across various platforms in terms of kernel coupling. Performance data was obtained using a variety of platforms, including the Cray X1, IBM p690 cluster, IBM SP3, and SGI Altix. The Prophesy performance analysis infrastructure was used to facilitate the creation of performance models to better understand the application. In addition, the performance data of additional applications was incorporated. Overall, an exploration of trends in the kernel coupling values across different applications and different system platforms were explored and are presented.

Using Genetic Algorithms for Adaptative Generation of Allocation Policies in Heterogeneous Distributed Systems
Wilson E. Lozano, Amado E. Pereira, John A. Sanabria and Wilson Rivera-Gallego, University of Puerto Rico - Mayaguez Campus, wilson.rivera@ece.uprm.edu

The problem of managing computational resources in heterogeneous distributed systems such as Grid-based systems is an NP Hard problem. As a result, a number of heuristic techniques have been developed to deal with the complexity of the problem.

A Grid-based testbed has been deployed at the University of Puerto Rico. This test-bed was used to obtain a benchmark of some known heuristic techniques for allocation of resources. A set of resource management techniques based on genetic algorithms have been explored and tested on this system and a systematic study of performance has been carried out. This analysis has provided new ideas to incorporate adaptive generation of allocation policies in heterogeneous distributed systems.

Technology and Access: Minority Serving Institutions Bridging the Divide
Debbie McMahon, Our Lady of the Lake University, ubergd@lake.ollusa.edu

This study examines how two Hispanic Serving Institutions in San Antonio, Texas, are using, conceptualizing and implementing technology, who is utilizing these technologies, and how computer policy is impacting the educational mission of the universities. Addressing “the perceived gap between those who have access to the latest technologies and those who do not” (NPEC), the U.S. Department of Education recognized and acted on the need to provide equity in higher education for people of color, with federal grants designed to reduce or eliminate the technology gap by providing funds to universities who qualify as Hispanic Serving Institutions. This study details changes that have taken place as a result of this funding, and what the momentum is to sustain the improvements made in education at these institutions with regard to technology. Interviews with faculty and key personnel indicate how program efforts to provide access and integration meet the needs of students and educational objectives of the institutions with regard to technological advancement and proficiency. This qualitative study explores the questions "What drives the technology initiatives to succeed and what does that success mean?"

Activities of Parallel Fibers on Cerebellar Purkinje Cells
Aaron Murray, University of Houston Downtown, murraya@gmail.com

In the cerebellum, Purkinje cells are very large, flask-shaped cells with huge dendrites extending upward to the molecular layer of the cerebellar cortex. Research on Purkinje cell models by Erik De Schutter and James M. Bower, in their paper, “Simulated responses of cerebellar Purkinje cells”, implied that Purkinje cell responses are independent of the dendritic location of the granule cell synaptic inputs. We systematically tested the validity of this theory with biophysically detailed models built using the Neuron modeling software package developed by Michael Hines and John W. Moore at the Department of Neurobiology, Duke University. An important addition was that of inhibitory collaterals between groups of Purkinje cells, a fact ignored in the original De Schutter and Bower study. The result of these studies indicated that collaterals should enhance small differences in excitation between Purkinje cells. These small differences could represent those in granule cell excitation, or even their location of excitation on the dendrites.

A Generic Interface for Communication between Story Understanding Systems and Knowledge Bases
Heather D. Pfeiffer, Nemecio R. Chavez, Jr. and Roger T. Hartley, New Mexico State University, hdp@cs.nmsu.edu

Presented is a story-understanding approach that uses world knowledge independent of the story at hand and independent of the structure or organization of the knowledge representation (KR). The world knowledge is stored and processed via a knowledge base (KB). This approach has been developed and tested, with "The Crow and The Pitcher" text story, using a prototype story-understanding system and two different KBs. Each knowledge base has different KR data structures and processing operations. We show that the generic interface between the story understanding system and each KB works equally well, and provides a platform where the relative efficiencies can be compared for the storage and retrieval operations. This poster discusses the outcome of comparing the two knowledge bases using this generic interface for story understanding on the given text passage, and suggests possible larger story problems for testing.

Simulations of Hand Motor Trajectories, Neuronal Electrical Activity, and Functional Magnetic Resonance Imaging (Fmri) Signal During Grasping Movements in Stroke
Antonio Ulloa and Marc Maier, ISERM, antonio.ulloa@alum.bu.edu

Functional magnetic resonance imaging (fMRI) studies in stroke patients have associated brain activity to the motor impairments resulting from stroke. However, the neural mechanisms responsible for impairments in grasping movements after stroke have not been systematically investigated. In the present work, we investigated the mechanisms responsible for motor impairment after stroke. We implemented a neural network model of the motor areas of the human brain and their interactions. The model simulates motor trajectories, neuronal electrical activity, and, after a transformation of the electrical activity, the fMRI signal that would correspond to that activity. The simulations show the effects of stroke as impairments of the grasping movement and as changes in neuronal electrical activity and fMRI signal. Our simulations agree with one experimental study on grasping and provide predictions of neuronal electrical activity and fMRI signal. Our model provides a mechanistic explanation of why stroke causes the observed motor impairments.

3D Finger Biometrics
Damon L. Woodard and Patrick J. Flynn, University of Notre Dame, dwoodard@nd.edu

As society becomes more electronically connected, the need for accurate automatic personal identification will increase. Biometrics addresses this need. Biometrics refers to the use of physiological and/or behavioral traits to automatically establish an individual’s identity. Examples of biometric identifiers currently in use include fingerprint, iris, and face. Our research investigates the use of the back surface of the hand, specifically the fingers, as a biometric identifier. We use 3D data from captured range images of the hand to compute a surface curvature based representation of the fingers’ surface. The normalized correlation coefficient is used to determine similarity during comparisons. Our experiments involve the use of a large database of images collected over time to assist in evaluating the stability of thometric identifier. We present our research results which indicate good performance for a first-of-its-kind technique and support our assertion that additional research in this area is warranted.