UTEP Distributed Computing Lab: Creation of a Pipeline for Minority Participation in Research
Project EIA-0325024 Report, 09/01/04 – 05/31/05 Research, Education and Outreach Activities
Patricia A. Nava, David H. Williams, and Joseph Pierluissi (pnava@ece.utep.edu, williams@ece.utep.edu, pier@utep.edu) Department of Electrical & Computer Engineering University of Texas at El Paso El Paso, TX 79968-0518 http://academics.utep.edu/Default.aspx?alias=academics.utep.edu/ee
ABSTRACT The UTEP Distributed Computing Lab (DCL) project, in addition to enriching the curriculum and facilitating interdisciplinary research, has assisted in the recruitment and support of underrepresented minorities, with the objective of creating a pipeline of students from high school through graduate school that are trained in the design, operation and applications of high-performance computers. The DCL project is in its second academic year. The 40-node cluster was built by students and the lab was established during our first year. The focus of this document is on this year’s accomplishments, which include: (1) fine-tuning of the cluster, which has resulted in 139 GFLOPS performance speed, and which serves as the hub for educational, research and outreach activities; (2) migration of the pilot research projects onto the cluster and evaluation of performance; (3) creation of cluster-based experiences for students in the curriculum; (4) engagement of graduate and undergraduate students in the DCL; and (5) conducting outreach to middle- and high-school students, particularly underrepresented minorities and women, to inspire their interest in CISE. Work to date has exceeded the second year goals, with advancement of the research, new collaborative efforts and the production of five student posters and 19 technical papers.
1. Introduction to UTEP
The University of Texas at El Paso (http://www.utep.edu) is a major regional university serving a rapidly growing, bi-national, bi-cultural population on the U.S.-Mexico border. El Paso, population approximately 700,000, is the fourth largest city in Texas and is one of the state’s fastest-growing metropolitan areas. The U.S.-Mexico border is a few hundred yards from the UTEP campus, and Ciudad Juarez, El Paso’s twin city across the Rio Grande, has a population of 1.3 million, creating a border community of around 2 million people. As the only major state university in Texas within 350 miles of El Paso, UTEP is central to providing educational opportunities and fostering the human and economic development of the region. El Paso is at the forefront of demographic trends that are rapidly changing the face of American society. The population of El Paso is estimated to be over 70 percent Hispanic, almost a quarter of El Paso’s population is first-generation immigrant, and over 50 percent of El Paso’s households speak Spanish as the language of preference. El Paso is also a young city, with a median age of 25 years, compared to 28 years for the state as a whole. Since the younger age brackets are more heavily Hispanic, public schools in El Paso County range from 75 to 98 percent Hispanic. Founded in 1914 as the Texas State School of Mines and Metallurgy, UTEP is the second-oldest academic component of the University of Texas system. UTEP's six colleges (business administration, education, engineering, liberal arts, nursing and health sciences and science) and the graduate school offer 60 bachelor's and 53 master's degrees. Doctoral degrees are offered in geological sciences, computer engineering, psychology, materials science and engineering, environmental science and engineering, and educational leadership and administration. With an annual research expenditure of $33 million, UTEP ranks among the top five universities (out of 36) in research expenditures in Texas, and 94 percent of its faculty hold a doctorate or equivalent in their field.
2. Introduction to the Project
The project consists of constructing a 40-CPU Beowulf Computer Cluster and its utilization in three interleaved goals: (1) to provide enrichment of educational activities in CISE; (2) to perform outreach to minority and female high-school students; and (3) to involve students in research and use the DCL as a catalyst to promote multidisciplinary research collaborations that can benefit from this form of computing. Individual research projects include neuro-fuzzy systems modeling, simulation, and applications; cluster architecture and networking; general distributed computation techniques; and enhancement of spatial resolution in electrocardiography through improvements in numerical methods. The work to date has directly supported goals (1) through (3) through multiple venues. The first goal is addressed by the formation of a “DCL Affinity Group,” support of students conducting research related to high performance computing, and inclusion of the cluster in the Spring 2005 Operating Systems course. The second goal is achieved through the DCL students’ participation in delivering the computer-centered module to middle- and highschool students in the Excellence in Technology, Engineering and Science (ExciTES) Summer Institute, sponsored by the Engineering Programs Office (EPO), as well as DCL tours and presentations provided to high-school students. Work on the third goal is in progress, but has been highly successful to date by mentoring students in research and making advances the pilot research projects, as detailed in the following section.
3. Accomplishments of the Past Year
The accomplishments of the project are described in terms of three categories. The first category addressed in section 3.1 is “Research,” which addresses the issues that have arisen with the cluster, performance research, and research in distributed solution of lineality (or linearity) space problems. This category also includes updates on the two pilot projects: Neuro-Fuzzy Simulation and Enhancement of Spatial Resolution in Electrocardiography. Section 3.2 addresses “Student Impact,” which reports on use of the cluster to enrich educational activities, DCL student achievements, Summer Camp outreach activities, high-school tours, and high-school presentations. Scholarly work, collaborative efforts and leveraged funding is reported in section 3.3.
3.1 Research
Systems Research: During the previous academic year, much care and thought went into the design of a 40-CPU Beowulf computer cluster. The student-constructed cluster was featured [1] in The Prospector, the campus newspaper during the Spring 2004 semester. In October, a follow-up article [2] entitled “Supercomputer Nears Completion,” detailed the status of the cluster and plans for its use in research and education. Nito Gumataotao, a graduate student within the ECE Department at UTEP, was primary architect of the final design that was then constructed by DCL students during three “build sessions.” Although the initial design consisted of a 32-CPU cluster, an equipment grant of three 3750 (24-port gigabit Ethernet communications) switches by Cisco Systems allowed the leveraging of funds in order to expand the design to a 40-CPU system in 20 compute nodes. The resulting cluster, named “Virgo” by the students who constructed it, forms the nucleus of the DCL. The front-end node [3] schedules work for the compute nodes and communicates with the outside world, while the compute nodes perform the desired computations. The nodes and switches form a private network of computers, which communicate only with each other. This private network, then, is externally accessible only through the front-end node. Thus, the architecture is distributed rather than parallel. Nito Gumataotao directed the construction of the cluster (a time-lapse movie of a built session is available for viewing at http://www.ece.utep.edu/research/webdcl/DCL-2/Templates/index.html) and currently administers its operation. The cluster employs Red Hat Linux [4] as the OS with ROCKS software [5] to manage and monitor the system. ROCKS is open source software distributed by the National Partnership for Advanced Computational Infrastructure (NPACI). ROCKS provides the mechanism to create a well-defined, secure, patched, system software for installation on all nodes. Monitoring is provided through “ganglia” which takes snapshots of node operation (CPU and memory usage, I/O, node up/down, etc.) approximately every five minutes or at the user’s discretion. ROCKS has already been found to be very useful software for management of the cluster. This year, the testing and debugging the Beowulf Cluster yielded several important results, which are reported with their associated tasks below. In addition, applications were run that are described in subsequent paragraphs. Major system tasks included:
Identifying and returning bad hardware modules. With 21 separate computing nodes, it is natural that some hardware problems would arise. We experienced several hardware failures, however, none were caused by student-assembly of the system. Failed hardware included two CPUs, an IDE disk drive, a SCSI disk drive, a CD-ROM drive, a power supply, and the frontend node motherboard. All units, except the CD-ROM drive were replaced under warranty by their respective manufacturers. Students in the DCL group assisted in diagnosing bad systems and tracing corrupt data, particularly that associated with the disk drives. The front-end motherboard caused random crashes which seemed unrelated until the problem was traced down. Since the bad motherboard was replaced, the frontend node has been running without problem. Obtaining device drivers for important elements of the system. Several device drivers were downloaded which were not included with the Red Hat Enterprise Linux 3.0 kernel, which came with “ROCKS”. Included were drivers for the CPU thermocouples and fan sensors, and for the NVidia graphics board. Several different kernels were tested before the current configuration was implemented. Benchmarking the system. System performance was tested using High Performance Linpack (HPL), the standard benchmark for high performance computers [6]. With various kernel configurations and different settings for HPL parameters, the performance has risen to 139 GFLOPS, using all compute nodes. At the present time, this performance ranks the cluster as the fastest computer system on campus (and perhaps fastest in the El Paso area).
Research on Distributed Solution of Lineality (or Linearity) Space Problems: One research project [7] investigated a distributed approach to solve lineality (or linearity) space (LS) problems using a Beowulf class cluster. A distributed approach is useful for these computations because substantial processing is required to solve LS problems with large cardinalities and a large number of dimensions. Our work showed that substantial speedups can be obtained by distributing the processing to multiple compute nodes. The LS solution has many important applications in engineering, science, and business, and includes a subset of solutions of the more general linear complementarity problem (LCP). A partial list includes convex quadratic programming solutions for the contact problem, the porous flow problem, the journal bearing problem, and free boundary problems, among others. Other applications of LCPs include the solution of equilibrium problems for market, capital stock, or traffic equilibrium, the Nash equilibrium problem and the n-person Nash-Cournot equilibrium problem. For this work, processing was distributed using copies of a C/MATLAB server executing on multiple compute nodes of a Beowulf cluster, and employed communications implemented with remote procedure calls (RPCs) incorporating external data representation (XDR). Up to four servers, one per compute node, were employed. With data sets of cardinalities of 500, 1000, 1500 or 2000 vectors, and dimensions of 4, 8, or 12 coordinates, the distributed computations were found to provide substantial speedups in comparison to execution on a single computer. For example, for a cardinality of 2000 vectors and 12 dimensions, processing time dropped from approximately 2600 seconds with one server, to approximately 700 seconds using processing distributed over four servers. Cluster Performance Research: One of the pilot research projects is neuro-fuzzy systems modeling, simulation, and applications. This project has made many advances, which are listed in the following paragraph, but the portion associated with performance is reported here. A study on the efficiency of distributing the code for a neural network simulation was reported in [8]. Subsequent research involves code for simulation of a Fuzzy Inference Neural Network, and its use on three standard benchmark problems: Detection of Breast Cancer, Diabetes Detection, and Automobile Efficiency Prediction. The facet associated with performance was a study looking at execution times and speed-up when distributing the application amongst different numbers of processors. The speed-up for these different applications is shown in Table 1, below. For these applications, distributed processing provides excellent speedup in comparison to a single computer. These results are slated for presentation [9] at the 2005 International Conference on Parallel and Distributed Processing Techniques and Applications (PDPTA 2005).
Table 1. Execution Speedup
No. of Processors 1 2 5 10 Breast Tumor Cancer Detection 1.00 1.38 2.68 8.85 Diabetes Detection 1.00 1.78 4.44 7.10 Automobile Efficiency (MPG) 1.00 1.86 3.12 5.82
Neuro-Fuzzy Simulation: The neuro-fuzzy research has spawned four lines of investigation: (1) fuzzy inference neural networks in conjunction with genetic algorithms [10, 11], and their application to standard classification problems (Diabetes Detection, Breast Tumor Cancer Detection, and Automobile Efficiency) [9]; (2) use of neural networks to extract knowledge for use in fuzzy systems [12], and their application to rat sleep classification [13]; (3) use of neural networks to classify respiratory diseases [14, 15]; and (4) use of neuro-fuzzy methods to classify human sleep via features extracted solely from an EEG signal [16, 17]. While the first involves collaboration with DCL PIs, the other three have collaborators that are other UTEP faculty or local physicians. All four lines of investigation have proven productive, and have generated topics for three master’s theses and one doctoral dissertation, as well as multiple publications, as referenced above. These four lines of investigation have, in turn, highlighted challenges or spawned other questions worthy of research, which will be examined in the following year. Enhancement of Spatial Resolution in Electrocardiography: This pilot project has generated several distinct research questions that are interleaved: (1) digitization of recorded electrocardiograms; (2) validated use of measured 12-Lead EKGs in the simulation of body surface potentials; and (3) an improved method for synchronization of electrocardiogram printouts. These investigations have led to collaboration with a biomedical engineer and a local cardiologist. A more detailed discussion of each of these studies is provided below. Digital Processing of Recorded Electrocardiograms: Digitization of EKG Printout software has been developed to convert an electrocardiogram into digital form for further analysis. Having digital EKG data facilitates a detailed analysis of properties of the EKG signals. Furthermore, it allows a digital vectorcardiogram to be obtained and used in simulation of body surface potentials. This research was presented as at the 22nd Annual Houston Conference on Biomedical Engineering Research [18]. Validated Use of Measured 12-Lead EKGs in the Simulation of Body Surface Potentials: This subproject aims to validate the use of measured 12-Lead EKG in the simulation of body surface potentials. This task is approached using the Digitization of EKG Printout software described in the previous paragraph. In order to simulate electrical activity of the human heart, the 12-Lead EKG in vector form, or vectorcardiogram, is needed. The technique used to obtain the vectorcardiogram was the Inverse Dower technique [19]. In addition, Frank’s XYZ transfer coefficient matrix for deriving the 12-Lead EKG [20] was used to validate the vectorcardiogram produced by the Inverse Dower technique. In order to produce an accurate vectorcardiogram, it is necessary to have a synchronized 12-Lead EKG recording. Since a digitized EKG printout does not provide a synchronized 12-Lead EKG, a synchronization method is developed. The method implemented in this research is synchronization by cross-correlation of an unsynchronized EKG with a synchronized EKG. This method transfers the synchronization of a 12-Lead EKG that was measured simultaneously to an unsynchronized-digitized EKG printout. This method proved to produce a more accurate vectorcardiogram because it reproduced the 12-Lead EKG more efficiently than other synchronization methods. The synchronized 12-Lead EKG can then be used to obtain the vectorcardiogram required for simulation of body surface potentials, as shown in Figure 1(a). Simulation of body surface potentials requires conversion the vectorcardiogram to a dipole moment. Validation of body surface potentials is achieved by comparing the measured precordial leads and the simulated precordial leads. The Normalized RMS error between the measured precordial lead and the simulated precordial leads is shown in Figure 1(b). Research results show that validation of simulated surface potentials is possible by comparing measured precordial leads and simulated precordial leads. To obtain an accurate diagnosis from a visualization of Body Surface Potentials, the visualization process must have the lowest error possible. This study was presented at the 22nd Annual Houston Conference on Biomedical Engineering Research [21].
Normalized RMS Error Between Measured Precordial Leads and Simulated Precordial Leads 40
Normalized RMS error (%)
30
20
10
0
V1
V2
V3 V4 V5 Precordial Recordings
V6
(a) (b) Figure 1.a) Surface potential simulation of a healthy patient used for validation of the use 12-Lead EKG in simulation of body surface potentials. b) Normalized RMS error between measured precordial leads and simulated precordial leads. Digitization and Synchronization Method for Electrocardiogram Printouts: This research is an improvement of the Digital Processing of Recorded Electrocardiograms study published in [18]. The improvement lies in the efficiency of the new code for the digitization of an EKG printout. This improved software is capable of synchronizing 24 EKG signals automatically and includes an improved method of synchronization than that presented in [21]. This study has been submitted for publication [22], and has spawned another collaboration on signal processing of the EKG [23].
3.2 Status of the Pipeline: Student Impact
Curricular use of the cluster: The cluster will be incorporated into several courses by August 2006. The first instance of its use for teaching occurred during the Spring 2005 semester. All students enrolled in EE 4374 (Operating Systems) were given a demonstration of the operation and use of the Beowulf Cluster. These activities included a demonstration of “ROCKS” [5] software, used to manage and monitor the operation of the cluster, and of applications that can take advantage of the processing power. In addition, an optional assignment was assigned which required the creation of a login account, and creation of software using Remote Procedure Calls (RPCs) to distribute matrix multiplication of 2 x 2000 and 2000 x 2 matrices using two compute nodes. DCL Affinity Group: The Distributed Computing Lab Affinity Group (DCL group) continues to meet once a week during fall, spring and summer semesters. The meetings this year have been used a venue for: discussion of lab issues (e.g. informal reports on status of replacement parts, outcomes of outreach activities, etc.) tutorials on professional skills (e.g. public speaking and effective presentation construction) exercising professional skills (students on the management team run the DCL Affinity Group meetings; all students present a 20-minute synopsis of their research, project or a current article; etc.) tutorials on cluster operation (e.g. command-line techniques for monitoring the system, etc.) team building activities (e.g. subgroups work on different requirements of the group: webpage team, library resources team, management team, publicity team, etc.) schedule and management planning for the summer camp (ExciTES) outreach activity
The numbers of students that have been participating in the DCL group are detailed in Table 2: DCL Student Summary.
Table 2: DCL Student Summary
Sponsored Students* Underrepresented Minorities Students by Degree Conference Participation
Academic Year
Total Students
Gender
Papers**
M
F
M
F
UG
MS PhD
2003-2004 2004-2005
24 29
19 21
15 18
9 11
12 16
5 11
8 15
15 11
1 3
2 13
Pos-ters
# Students Participating in Outreach Activities
1 2
12 10***
NOTES: * – Number of students sponsored, directly or with leveraged funding. ** – Student co-author, preparation of slides and/or conference presentation. *** – Does not include 2005 summer camps, which will take place in July 2005.
Outreach (ExciTES Summer Institute): The 2005 Summer Camps will take place in July. We have projected the involvement of 70 local middle- to high-school students for this year. In lieu of results for the 2005 Summer Camps, the results for the 2004 Summer Camps are reviewed. During the first year, the outreach plan leveraged the DCL efforts with existing programs and offices in order to ensure the effectiveness of the outreach. The ExciTES Summer Institute activities consisted of a DCL module and other science and engineering modules, as well as participation in events sponsored by the College of Engineering Programs Office (EPO). Three week-long summer camps were conducted. The outreach is focused on local students ranging from rising 8th graders to rising juniors, and the first year’s activity reached 64 students. DCL students delivered the module entitled “Building a PC,” developed by the PIs with the assistance of DCL group members. This module consisted of lecture time, providing an introduction to the CISE field as well as this project and computers in general. This was followed by a hands-on session where the ExciTES participants constructed a PC. Then, campers used it to run an application upon successful construction. This activity proved interesting to the outreach participants and was met with enthusiasm by the DCL (college) students. Over 12 DCL students volunteered to mentor/assist in this PC construction event. The assessment showed that the second session participants experienced a 56% increase in general awareness of, and interest in, CISE studies and careers. The third session experienced a 20% increase in the same. The instrument used for assessment included area for comments, which were very telling about the effectiveness of the camp. In summary: 64 high school students participated in the camps Questionnaires show high level of interest in DCL module DCL students served as role models DCL students participated in developing materials and delivering the modules Campers’ comments, as well as assessment results, indicate the module is effective at raising levels of awareness and interest Outreach (UTEP Engineering EXPO and high school tours): In addition to exposing high school and middle school students to high performance computing through the ExciTES summer institute, operation of the Virgo cluster was demonstrated to several groups of students and their teachers during the Engineering EXPO program in February. The groups were from various high schools in the El Paso metropolitan area with students that were freshmen through seniors. Each group was briefed on the overall structure of the cluster hardware and software, shown the machine in operation, and given demonstrations of applications, both graphical and computational which are sped up by distributed processing. Emphasis was placed on the fact that Virgo was constructed and is operated by students, and as a member of the DCL group, they too could have hands-on access to the cluster. Several student members of
the DCL group assisted in the presentations. This activity generated such interest that an “EXPO extra” was scheduled for the following weeks. During the “extra” sessions, our students hosted a total of 165 students from local schools such as Chapin and Parkland. Outreach (Chapin High School Engineering Week): Chapin High School was established three years ago. It is an engineering magnet school, and has collaborated with UTEP to celebrate Engineering Week by hosting speakers and presentations about engineering. This year, six DCL students participated in three different sessions, making presentations on “UTEP’s Distributed Computing Lab,” “High Performance Computing” and “3-D Heart Visualization.” Approximately 120 students attended these presentations.
3.3 Publications, Associated Projects and Funding Efforts
The DCL has provided a venue for collaboration and publication efforts that have been very effective. Specifically, this work has led to two journal submissions [7, 11], the publication of eleven conference papers [8, 9, 13, 15, 17, 22-27], two conference posters [18, 21], three Master’s theses [12, 14, 16], and a doctoral dissertation [10]. The grant has also provided the PIs and participating students the opportunity to collaborate with three physicians, two biomedical engineers, a biologist and a sleep researcher. This collaboration has led to the submission of proposals to various agencies, as shown in the bulleted list below. In addition, the grant has been leveraged with funding from other sources. For example: “Request for Three Cisco Gigabit Switches for a 32 CPU Beowulf Cluster,” Cisco Systems, Inc. (Higher Education Equipment Grant), PIs: Williams (Lead), Nava, Pierluissi, 2/1/04-1/31/05, $22,305 (awarded). “Algorithm Partitioning and Mapping,” Subproposal to UTEP’s NSF-CREST Submission, PIs: Williams (Lead), Nava, Diong, Chiu, Pan, 9/1/04-1/31/09, $478,280 requested (declined). “Impulse Oscillometric Evaluation of the Effect of Air Quality on Respiratory Function in Normal and Asthmatic Anglo and Hispanic Children at the Border,” Subproposal to UTEP’s National Institute of Environmental Health Sciences – Advanced Research Cooperation for Environmental Health (NIEHS–ARCH) Submission, PIs: Nazeran (Lead), Nava, Diong, 9/1/05- 8/31/08, $296,754 requested, (decision pending). One student funded by UTEP Graduate School One student funded by the PI’s Teaching Award One student funded by UTEP Department of Electrical and Computer Engineering One student funded by another NSF-MII grant, “Graduate Education for Minority Students in Computer Science and Engineering: Extending the Pipeline”
4. Immediate Impact
Our second year has proven productive, with successful research pilot projects, multiple publications, promising collaborations, involvement of women and under-represented minorities in our DCL Affinity Group, support of those students, and exciting outreach events that have potentially laid the groundwork for more students in the CISE pipeline.
5. Project Outcome
The project has successfully met this year’s goals, as detailed in section 3, “Accomplishments of the Past Year.” In summary, our goals are: (1) to provide enrichment of educational activities in CISE; (2) to perform specialized outreach to minority and female high-school students; and (3) to involve students in research and use the DCL as a catalyst to promote multi-disciplinary research collaborations that can benefit from this form of computing. Goal (1) has been met by involving the DCL students in the following activities: integration of the cluster into one course in the curriculum; “hands-on” sessions with the undergraduates;
delivery of modules on MPI techniques, and mentoring of undergraduates by graduate students; participation in weekly affinity meetings, with one presentation per student per semester; and writing, preparing and delivering technical papers and posters at conferences.
Goal (2) has been met by conducting three ExciTES camps during the months of June and July, 2004, reaching a total of 64 middle- to high-school students. It will continue with ExciTES camps projected to reach 70 middle- to high-school students, planned for July 2005. Goal (3) has been met by: success in research pilot projects of systems research, neuro-fuzzy research and electrocardiography enhancement research; completion of 19 scholarly works involving PIs and DCL students, with students being the first author on 18 of those; collaborative research with electrical engineers, biomedical engineers, physicians, and biologists; submission of at least three other proposals to fund collaborative efforts. The project’s outcomes, which fit with our goals, would not have been possible without the grant. Indeed, all three explicit goals listed dovetail to enhance the overarching goal of creating a pipeline, from high-school to graduate school, of students interested in CISE studies. The award has already made possible at least several breakthroughs that would not have otherwise been practical: establishment of a distributed computing lab; increasing awareness of CISE studies and careers amongst pre-college students; increasing the number of college students interested in distributed computing; and providing incentive for faculty with other expertise to join a collaborative research effort.
6. References
NOTE: student authors designated with an *. [1] Trejo, Erica, “Student[sic] Build Supercomputer from Scrach[sic],” The Prospector, University of Texas at El Paso, April 14, 2004, p 6. [2] Hall, John, “Supercomputer nears completion,” The Prospector, University of Texas at El Paso, October 6, 2004, p 8. [3] http://www.beowulf.org [4] http://www.redhat.org [5] http://rocks.npaci.edu/Rocks/ [6] www.top500.org [7] Caire*, M. E., F. J. Lopez, and D. H. Williams, “Distributed Identification of the Lineality Space of a Cone”, submitted to the International Journal of Computers and Applications, 2005. [8] Sagarnaga*, Miguel A., David H. Williams, and Patricia A. Nava, “Distributed Processing of a Neural Network Application”, Proceedings of the 2004 International Conference on Parallel and Distributed Processing Techniques and Applications, pp. 119-124 June 21-24, 2004, Las Vegas, NV. [9] Cruz-Cano*, Raul, Patricia A. Nava, and David H. Williams, “A Parallel Genetic Algorithm for Fuzzy Inference Neural Networks,” accepted for the 2005 International Conference on Parallel and Distributed Processing Techniques and Applications, June 27-30, 2005, Las Vegas, NV. [10] Cruz-Cano*, R., Creation and Optimization of Fuzzy Inference Neural Networks, Ph.D. Dissertation, University of Texas at El Paso, May 2005. [11] Cruz-Cano*, R., P.A. Nava, and D.H. Williams, “Fuzzy Inference Neural Networks: A Novel and Efficient Method for Optimization,” submitted to the Journal of Intelligent and Fuzzy Systems, May 2005. [12] MartinDelCampo*, E., Design of Fuzzy Systems using Knowledge Extracted via Neural Networks, MSEE Thesis, University of Texas at El Paso, August 2004. [13] Cruz-Cano*, R., E. MartinDelCampo*, P.A. Nava, and R. Cabeza, “Fuzzy Rule Extraction and Optimization for Rat Sleep-Stage Classification,” accepted for the 2005 International Conference on Parallel and Distributed Processing Techniques and Applications, June 27-30, 2005, Las Vegas, NV. [14] Barua*, M., Classification of Pulmonary Diseases Using Artificial Neural Networks, MSEE Thesis, University of Texas at El Paso, December 2004.
[15] Barua*, M., H. Nazeran, P. Nava, and V. Granda*, “Classification of Pulmonary Diseases using Artificial Neural Networks,” Proceedings Of ANNIE 2004, November 2004. [16] Barragan*, J., Neuro-Fuzzy Sleep Classification using Features Extracted from EEG Signals, MSEE Thesis, April 2005. [17] MartinDelCampo*, E., E. Estrada*, J. Barragan*, P. Nava, and H. Nazeran, “Automatic Sleep Stage Classification Utilizing Fuzzy Logic and Neural Network with the EEG Signal,” Proc. Of ANNIE 2004, November 2004. [18] Sevilla*, D., E. Morales*, J.H. Pierluissi and Z. Abedin, “Digital Processing of Recorded Electrocardiograms,” Proceedings of the 22nd Annual Houston Conference on Biomedical Engineering Research, 2005. [19] Edenbrandt, Lars and Olle Pahlm, “Vectorcardiogram synthesized from a 12-lead ECG: superiority of the inverse dower matrix,” Journal of Electrocardiology, vol. 21, issue 4, pp. 361-367, 1988. [20] Frank, E., “An accurate, clinically practical system for spatial vectorcadiography,” Circulation, vol.13, pp.737749, 1956. [21] Morales*, E.,D. Sevilla*, J.H. Pierluissi, and Z. Abedin, “Validated Use of Measured 12-lead EKGs in the Simulation of Body Surface Potential”, Proceedings of the 22nd Annual Houston Conference on Biomedical Engineering Research, 2005. [22] Morales*, E., D. Sevilla*, J.H. Pierluissi, and Z. Abedin, “Digitalization and Synchronization Method for Electrocardiogram Printouts”, accepted for the 27th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Shanhai, China, 2005. [23] von Borries, R., J.H. Pierluissi, and H. Nazeran, “Wavelet Transform-based ECG Baseline Drift Removal for Body Surface Potential Mapping”, accepted for the 27th International Conference of the IEEE Engineering in Medicine and Biology Society, Shanhai, China, 2005. [24] Estrada*, E., H. Nazeran, P. Nava, K Behbehani, J. Burk, and E. Lucas, “Itakura Distance: A Useful Similarity Measure between EEG and EOG Signals in Computer-aided Classification of Sleep Stages,” submitted to the 26th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2005. [25] Joshi*, K., S. De La Cruz*, B. Diong, D. Williams and P. Nava, “A Neural Network Implementation of Dualfrequency Output Control for Multilevel Inverters,” Proc. Of the 2004 ComPEL Conference, 2004. [26] Barua*, M., H. Nazeran, P. Nava, V. Granda*, and K. Behbehani, “Classification of Pulmonary Diseases by an Artificial Neural Network, Based on Measurements from the Impulse Oscillometry System,” Proc. Of the 26th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2004. [27] Woo*, T., B. Diong, L. Mansfield, H. Nazeran, M. Goldman, and P. Nava, “A Comparison of Various Respiratory System Models Based on Parameter Estimates From Impulse Oscillometry Data,” Proc. Of the 26th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2004.