University of Nevada, Reno - Department of Computer Science & Engineering [Past Colloquia]

CSE Senior Projects 2009

CSE annual open house and workshop

Location: REL 109-110

Friday, May 1st, 2009 at 09:30a.m.

2009 Computer Science and Engineering Senior Projects Workshop

9:30 am
Project 1 DOCS (Document Organization Computing Solutions)
A web application for organizing physical files in offices (e.g., law offices)
Students Darren Johnson, Kellen Dale, Joel Martinez, Vic Van Horn

Advisors Mr. Wes Hays, Mr. John Dell, CASAT Reno

10:00 am
Project 2 Slot Tracker
A system for Bally Technologies that wirelessly tracks slot machines on casino floors
Students Katie Ceglia, David Keele, Michael Picerno, Ben Seelbinder
Advisor Mr. Ryan Ruppert, Bally Technologies

10:30 am
Project 3 Muse – A Music Theory Game
An interactive edutainment software that helps learning music theory
Students Ben Brown, Josh Greenspan, Wesley Lee
Advisor Dr. Eelke Folmer, UNR

11:00 am
Project 4 GELS (General Electric Licensing System)
A web application for managing General Electric’s software licensing
Students Miran Kim, Xiaolu (Lulu) Zhang, Victor Ivanov
Advisor Mr. Ken Ceglia, General Electric

11:30 am – 1:00 pm CSE Advisory Board meeting & lunch

1:00 pm Posters session: research, game development, and senior projects

1:30 pm
Project 5 Wield–CAVE (Wireless Ergonomic Lightweight Device – for use in CAVE)
An interaction glove for immersive virtual reality simulation and training
Students Joshua Hegie, Andrew Kimmel, Kelvin Parian
Advisor Dr. Frederick C. Harris, Jr., UNR

2:00 pm
Project 6 HEDC (Human Error Detection and Correction)
A human error detection and correction navigation system implemented on a Segway robot
Students Trevor Bullock, Arthur LeVesque, Ryan Luna
Advisor Dr. Kostas Bekris, UNR

2:30 pm
Project 7 IRIS (Image Recognition Inventory System)
An image-based medical instruments inventory system for hospitals in Africa
Students Matt Whipple, Cody White (with Fabien Pitetti, Vincent Roydor)
Advisors Dr. Philippe Dugerdil and Mr. Francois Wahl, Geneva, Switzerland

3:00 pm
Project 8 Standalone Hard Drive Manager
A standalone hard disk test manager system for PC-Doctor
Students Amanda Sou, Javier Garcia, Curtis Richardson
Advisors Mr. Ken Sheppard and Mr. Hector Urtubia, PC-Doctor, Inc, Reno, NV

3:30 pm
Project 9: iPack – A UNR iPhone Application
A UNR information and navigation system for the iPhone
Students A.J. Hernandez, Chris Maloney, Jeff Naruchitparames
Advisor Mr. Brian Westphal, CliqCliq

CSE Department, UNR

Ms. Donna Djordjevich, Sandia National Labs Mr. James Kosta, 3G Studios Dr. Michael Mateas, UC Santa Cruz

IGT-UNR Engineering Symposium

Location: Mathewson IGT Knowledge Center Room 124 April 17, 2009

Friday, April 17th, 2009 at 09:30a.m.

Sponsored and organized by the IGT and CSE Department, UNR. The event will run from 9:30 am to 3:30 pm. Short tours of the Mathewson-IGT Knowledge Center are included and lunch will be provided based on availability.

Emerging Technologies in Games and Gaming

Agenda
9:00 Registration opens
9:30 Morning Tour - Mathewson IGT Knowledge Center
10:00 Welcome
10:10 Speaker #1 – James Kosta
11:15 Panel
12:00 Luncheon – MIKC 201 Rotunda
12:45 Afternoon Tour - Mathewson IGT Knowledge Center
1:15pm Speaker #2 – Michael Mateas
2:20pm Speaker #3 – Donna Djordjevich
3:20pm Event Ends

James Kosta – “Iterative Game Design: Analyzing Usage as Part of the Production Cycle”
Mr. Kosta is the CEO and founder of 3G Studios. Prior to entering the video game industry, he led product and software design projects at Disney, Dell, Microsoft, Hewlett-Packard, Compaq and Digital Equipment Corporation (DEC). With video game development budgets growing exponentially, it is important to have a strong pre-production methodology and constant play testing. Mr. Kosta will teach techniques to help strengthen your processes while allowing you more creative freedom to polish and refine.

“The Convergence of Casino Gaming and Computer Games: Potentials and Barriers” Join the following esteemed panelists for a lively and informational chat about how video and non-wagering game technologies are converging with traditional casino game development: Larry Dailey, UNR School of Journalism; Jim Hunt, Bally Technologies; James Kosta, 3G Studios; Tim Page, 5000ft; and James Vasquez, IGT.

Michael Mateas – “Artificial Intelligence for Video Games”
Mr. Mateas is a computer science department faculty member at UC Santa Cruz. He helped launch the game design program there, offering the first such degree in the UC system. His research in AI-based art and entertainment combines science, engineering and design into an integrated practice that pushes the boundaries of the conceivable and possible in games and other interactive art forms.

Donna Djordjevich – “The ‘Ground Truth’ First Responder Training Game”
Ms. Djordjevich is a senior member of the technical staff under the Homeland Security and Defense Systems Center at Sandia National Laboratories. She holds a master’s in Computer Science from USC. She currently leads the development of Sandia’s “Ground Truth” training video game. This platform immerses first responders in an interactive gaming environment depicting high-risk, high-threat situations experienced from the comfort of their desk.

IGT and CSE Department, UNR

Dr. Pavel Solin

Department of Mathematics and Statistics, UNR

Biography

Pavel Solin is an associate professor in applied mathematics at UNR. He received a Ph.D. in Mathematical and Computer Modeling from the Charles University, Prague. He is the recipient of major awards for young scientists in the Czech Republic such as the Bolzano, Hlavka and Babuska awards, and the Presidential award for young scientists. Dr. Solin held postdoctoral positions at the Kepler University (Austria), UT Austin and Rice University. His group is developing novel high-fidelity computational methods for partial differential equations (PDE) and integral equations. Dr. Solin wrote three research monographs on these topics. He is the organizer of bi-annual international conferences "European Seminar on Coupled Problems (ESCO)" and "Finite Element Methods in Engineering and Science (FEMTEC)" that take place in Europe and the U.S., respectively. Dr. Solin serves on editorial boards of several international journals, leads an international research group, and has collaborators in numerous European countries and the U.S.

Location: SEM-347

Friday, April 3rd, 2009 at Noon

Sponsored and organized by the CSE/EBME/IEEE

Adaptive Finite Element Methods: From PDE to Image Compression

Partial Differential Equations (PDE) are used to model many important processes in the nature such as atmospherical processes, flow of liquids, deformation of solid objects, geomagnetic and electromagnetic fields, transfer of heat, and others, but also to model developments on financial markets or the structure of molecules. These equations cannot be solved exactly and thus we approximate them numerically using high-performance computing. There are various numerical methods for PDE but a prominent role plays the Finite Element Method (FEM) which is used heavily in many engineering and scientific disciplines. We will give examples, mention some problems, limitations, and dogmas related to modern computational methods, and describe what we are doing in order to overcome them. We will also present the open source project Hermes (http://hpfem.org), a C++/Python library for rapid prototyping of high-fidelity solvers for PDE and PDE systems developed by our group. We will describe the goals of the project, how we want to achieve them, and what makes Hermes different from other PDE solvers and libraries.

UNR CSE Host

Dr. George Bebis

Dr. Sven Koenig

University of Southern California

Biography

In this talk, I will give an overview of ongoing research at USC on games and then talk about how research on artificial intelligence and robotics can inspire research on games. In particular, I will talk about our research on designing pinball games and on path planning for video games. In the context of any-angle path planning on grids, I will present Theta*, a variant of A* that propagates information along grid edges without constraining the paths to grid edges, and demonstrate that Theta* is simple, fast and finds short and realistic looking paths. In the context of moving-target search, I will present MT-Adaptive A*, an incremental variant of A* that updates the heuristics between searches, and demonstrate that MT-Adaptive A* is faster than isolated A* searches and, in many situations, also D* Lite, a state-of-the-art incremental variant of A*. This is joint work with K. Daniel, D. Earl, A. Felner, M. Likhachev, A. Nash, X. Sun, D. Wong and F. Zyda.

Location: UNR Knowledge Centrum, Auditorium 124

Monday, March 9th, 2009 at 4:00p.m.

Sponsored and organized by the CSE/EBME/IEEE

On Robotics and Games

Sven Koenig is an associate professor in computer science at the University of Southern California. He is the recipient of an NSF CAREER award, an IBM Faculty Partnership Award, a Charles Lee Powell Foundation Award, a Raytheon Faculty Fellowship Award, and an ACM Recognition of Service Award, among others. He co-founded Robotics: Science and Systems, was conference co-chair of the 2002 Symposium on Abstraction, Reformulation, and Approximation, conference co-chair of the 2004 International Conference on Automated Planning and Scheduling, program co-chair of the 2005 International Joint Conference on Autonomous Agents and Multi-Agent Systems and program co-chair of the 2007 and 2008 AAAI Nectar programs. Additional information about Sven can be found on his webpages: idm-lab.org.

UNR CSE Host

Dr. Kostas Bekris

Mr. Todd G. Shipley

President and CEO, Vere Software

Biography

Todd G. Shipley, CFE, CFCE, President and Chief Executive Officer of Vere Software. Mr. Shipley is a retired City of Reno Police Detective Sergeant, where he started and managed Nevadas first cybercrime unit. Prior to designing WebCase and starting Vere Software he was the Director of Systems Security and High Tech Crime Prevention Training for SEARCH, The National Consortium for Justice Information and Statistics. He oversaw a national program that provided expert technical assistance and training to local, state, and federal justice agencies on successfully conducting high-technology computer crimes investigations. In this position he was also the manager of the National Criminal Justice Computer Laboratory and Training Center. Mr. Shipley has also achieved his Certified Forensics Computer Examiner (CFCE) certification from the International Association of Computer Forensic Specialists and is currently the 1st Vice-President of the High Technology Crime Investigation Association. He has authored works regarding cybercrime and speaks nationally on cybercrime investigations.

Location: SEM-347

Friday, March 6th, 2009 at Noon

Sponsored and organized by the CSE/EBME/IEEE

An introduction to Digital Forensics

This session is designed to impart to the participants the basics of digital evidence and its use in the legal system. This session will cover the current status of digital forensic evidence from both the scientific and legal perspectives including a review of cybercrimes their background and how they affect the public and the law.

UNR CSE Host

Dr. George Bebis

Dr. Michael S. Branicky, Program Director

National Science Foundation

Biography

Dr. Branicky is currently a Program Director at the National Science Foundation (NSF) in the Computer & Network Systems (CNS) Division of the Computer & Information Science & Engineering (CISE) Directorate. He is involved with the following NSF programs: Cyber-Physical Systems (CPS), Computer Systems Research (CSR; core research and CAREER awards), Computing Research Infrastructure (CRI), Cyber-enabled Discovery and Innovation (CDI), Expeditions in Computing (Expeditions), and Science and Technology Centers (STC). Dr. Branicky is also Professor of Electrical Engineering and Computer Science at Case Western Reserve University, Cleveland, OH. He received the B.S. (1987) and M.S. (1990) in Electrical Engineering and Applied Physics from Case and the Sc.D. in Electrical Engineering and Computer Science from the Massachusetts Institute of Technology (1995). He has held research positions at MIT`s AI Lab, Wright-Patterson AFB, NASA Ames, Siemens Corporate Research (Munich), and Lund Institute of Technology`s Dept. of Automatic Control. Research interests include hybrid systems, intelligent control, and learning, with applications to robotics, manufacturing, control over networks, and biology.

Location: Auditorium 124, UNR Knowledge Center

Friday, February 6th, 2009 at 1:30p.m.

Sponsored and organized by the CSE/EBME/IEEE

Cyber-Physical Systems Initiative: NSF & Beyond

The National Science Foundation has a new initiative (Solicitation 08-611; proposal deadline 02/27/2009) to support transformative research on the theory and applications of Cyber-Physical Systems (CPS). The term "cyber" refers to computation, communication, and control. The CPS challenge is motivated by systems in which cyber and physical elements are deeply integrated and networked at all scales. Research advances in CPS promise to transform our world by delivering autonomous cars and aircraft, smart buildings attached to an intelligent power grid, and new medical/assistive technologies. Delivering such systems in a way reliable enough to "bet our lives on" represents significant scientific and technical challenges. Solutions will enable applications like those above, as well as impact U.S. security and competitiveness in a number of critical sectors, including aerospace, automotive, chemical production, civil infrastructure, energy, healthcare, manufacturing, materials, and transportation. This talk will present the motivation and vision behind the NSF`s CPS initiative and summarize the program details. It will also suggest ways in which researchers can engage in the nascent CPS community, and how CPS research could impact its constitutive disciplines. Other opportunities at NSF will also be overviewed as time permits.

UNR CSE Host

Dr. Kostas Bekris

Dr. Lee Barford

Agilent Laboratories, Santa Clara, CA

Biography

Lee Barford received a PhD in Computer Science from Cornell in 1987. From there, he went to Hewlett Packard Laboratories and Agilent Laboratories, where he has worked primarily on software for use by engineers in other disciplines. One product resulting from his research, Fault Detective Test Analyzer, an automatic fault isolation modeling tool, won the Electronic Design News (EDN) award for software innovation of the year 2002. Lee has given invited talks at conferences with topics as disparate as model-based diagnosis and semiconductor test. He currently leads a group investigating how the software architecture of scientific and engineering instrumentation and sensor systems will need to change in response to the pervasiveness of parallel computing.

Location: SEM-347

Friday, November 14th, 2008 at Noon

Sponsored and organized by the CSE/EBME/IEEE

Bridging Atoms and Actionable Information

When someone pays for the physical world to be sensed, they are rarely primarily interested in the raw or calibrated sensed physical quantities. Instead, what is typically desired is actionable information about the physical situation. That is, what is wanted is information that enables decision making, such as when and what to repair in a system, whether a product is of adequate quality, or what steps to take to improve traffic throughput. Typically, multiple levels of analysis are required to translate sensor outputs into actionable information, starting with traditional signal processing and often continuing with higher levels of abstraction, applying artificial intelligence methods. Examples will be described from telecommunications, transportation, and manufacturing.

UNR EBME Hosts

Dr. Yantao Shen and Dr. Xiaoshan Zhu

Dr. Jim Whitehead

University of California, Santa Cruz

Biography

Jim Whitehead is an Associate Professor of Computer Science at the University of California, Santa Cruz, where he led the creation of an undergraduate degree program, the BS Computer Science: Computer Game Design. Jim is the General Chair of the 2009 Foundations of Digital Games conference. His research interests span software engineering and computer games; the intersection of these is the study of automatic generation of computer game levels. Jim also studies the software bug prediction and software evolution. He received his PhD in Information and Computer Science from UC Irvine in 2000, under his advisor Richard N. Taylor.

Location: SEM-347

Friday, October 24th, 2008 at Noon

Sponsored and organized by the CSE/EBE/IEEE

Computer Games as an Academic Discipline

Computer games are increasingly the focus of serious academic inquiry. An inherently computational medium, games have spawned a wide range of study spanning disciplines from the arts and humanities to computer science. This is leading to the formation of computer games as an academic discipline. This talk explores what characteristics a medium must have such that it can sustain serious long-term inquiry. How a new academic discipline is legitimized is discussed, focusing on how different institutions can contribute legitimacy. The current state of disciplinary formation of computer games is discussed, with recommendations for future action. The talk ends with a sketch of several societal-scale interventions that could result from academic engagement with computer games.

UNR CSE HOST

Dr. Sushil Louis

Dr.Kenneth A. Marko

Chief Scientist ETAS Inc.

Biography

Dr. Ken Marko is a physicist by training who has worked in several different disciplines. As a staff member of the Physics Department at the University of Michigan, he carried out fundamental research on the behavior of matter-antimatter systems to test if theoretical predictions about these ephemeral systems could be confirmed by experimental facts. He then joined Ford Motor Company as a scientist in the Physics Department where he investigated the chemical kinetics and the fluid dynamics of combustion inside prototype engines using a combination of non-linear coherent Raman spectroscopy and pulsed laser flow field imaging with early versions of image intensified TV cameras. He later began work, as Team Leader for Advanced Diagnostics, on development of diagnostic methods for complex systems based both on an understanding of the fundamental physics of system operation and the use of machine learning (neural networks) to produce extremely efficient and accurate diagnostic algorithms which are currently in production use worldwide for OBD. He has most recently begun to develop methods for prognostics, as well as diagnostics, based on embedded software agents. He has served as a Six Sigma Black Belt Champion successfully completing projects including an effort to extend battery life for hybrid vehicles. He is one of the organizers of the Society of Automotive Engineers yearly TopTecs on On-Board Diagnostics in the US and Europe and was the General Chairman of the International Joint Conference on Neural Networks in 2001. He holds a Bachelor of Science degree from MIT and an M.S. and Ph.D. in Physics from the University of Michigan. He is the author of more than 70 publications and holds 18 patents.

Location: HREL-109

Wednesday, August 27th, 2008 at 4:00p.m.

Sponsored and organized by the CSE/EBE/IEEE

IMMUNE SYSTEM ENGINEERING New Methods of Machine Learning Applied to Control, Diagnostics and Prognostics of Automotive Systems

In the automobile industry, which includes passenger vehicles as well as truck and off-road machines, the rising price of fuel and the need to limit emissions have created major challenges for system designers and production engineers. Additionally, the internationalization of the industry has substantially increased competitive market pressures to produce unique and superior systems. System complexity has reached levels which make standard design and validation processes obsolete and unable to provide competitive products at affordable cost. In complex systems, the problems of system design, design validation (i.e. testing), optimal control, diagnostics and prognostics for condition based maintenance will require major innovations before ideas on the drawing boards can be realized in affordable products. In fact, government regulations require many new systems to provide their own internal diagnostics in order to qualify for production and public sale. We have developed a concept called Immune System Engineering to address many of these issues. This concept is based upon the notion that diagnostics has to be embedded into complex systems from the outset, and that these embedded diagnostics can be used to improve designs, expedite validation, improve performance, sustain operation and enable more efficient repairs. The basis for this capability is the development of high fidelity models of complex systems, which are often highly nonlinear dynamic systems which are not adequately simulated by first principle models. The basis for the machine learning models is the use of dynamic neural networks trained by complex, but powerful new learning algorithms, which can mimic many forms of complex systems. I will discuss the development and application of this strategy to a number of challenging problems in the industry and provide demonstrations that the methods we propose are effective and realizable in mass produced systems.

UNR CSE HOST

Dr. Bobby D. Bryant

Mehmet Gunes

University of Texas at Dallas

Biography

Mehmet Gunes is currently at the University of Texas at Dallas receiving his Doctor of Philosophy in Computer Science. He received his Masters of Science in Computer Science & Engineering from Southern Methodist University in Dallas and his bachelor or Science in Computer Science and Engineering from Isik University in Istanbul, Turkey. His research interests include network measurements, network protocols, network security, complex networks, graph sampling, graph data mining, biological network analysis.

Location: SEM 201

Tuesday, May 27th, 2008 at 11:00a.m.

Sponsored and organized by the CSE

Internet Topology Discovery

Understanding the topological characteristics of the Internet has been an important research issue. This understanding is not simply an intellectual curiosity but also a necessity to better design, implement, and operate the underlying network technologies, protocols, and services. It is also a requirement for protection of critical infrastructures that depend on the Internet. In general, Internet topology measurement studies involve three phases: topology collection, topology construction, and topology analysis. The first step is to collect topology information using various measurement approaches. In topology construction step, the raw data obtained in the topology collection phase is converted into a corresponding topology map.

Finally, constructed network topologies are analyzed. This talk presents our work on topology construction in the context of router-level Internet topology measurement studies to discover Internet topologies more accurately.

UNR CSE Host

Dr. Dwight Egbert

Dr. Konstantinos Bekris

Department of Computer Science, Rice University

Location: SEM 201

Wednesday, April 9th, 2008 at 11:00a.m.

Sponsored and organized by the CSE

Vehicular Localization in Urban Environments using a Wireless Mesh Network

Locating a device's position using information from wireless communication has attracted a lot of attention in several research communities because it provides with a multitude of location-based services. This report focuses on the instance of this problem that involves such a device carried by a vehicle driven in an urban environment that is covered by a wireless mesh network. The proposed approach uses data gathered by a location-aware vehicular user of the network to build a model of expected observations in the covered region. Subsequently, vehicles can be localized online using only information from a wireless adapter and by applying a bayesian, probabilistic inference method. The localization accuracy is at the block level, allowing applications, such as specialized web search, navigation, social networking or high-level activity inferencing.

UNR CSE Host

Dr. George Bebis

Dr. Biplab Sikdar

Rensselaer Polytechnic Institute

Biography

Biplab Sikdar is currently an Associate Professor in the Department of Electrical, Computer and Systems Engineering of Rensselaer Polytechnic Institute, Troy, NY, USA. He received the B. Tech degree in electronics and communication engineering from North Eastern Hill University, Shillong, India, the M. Tech degree in electrical engineering from Indian Institute of Technology, Kanpur and Ph.D in electrical engineering from Rensselaer Polytechnic Institute, Troy, NY, USA in 1996, 1998 and 2001, respectively. His research interests include wireless MAC protocols, network routing and transport protocols, performance evaluation and queueing theory.

Location: MS 321

Friday, April 4th, 2008 at 11:00a.m.

Sponsored and organized by the CSE/EBE/IEEE

Energy Efficient Transmission Strategies for Body Sensor Networks with Energy Harvesting

The limitations caused by current battery technology is one of the main obstacles in the path of widespread deployment of Body Sensor Networks (BSNs), which have the potential to significantly benefit numerous medical and non-medical applications and services. This presentation will describe our work on addressing the problem of developing energy efficient transmission strategies for body area sensor networks with energy harvesting capabilities. Taking into account the energy harvesting capabilities of the sensor, decision policies are developed to determine the transmission mode to use at a given instant of time in order to maximize the quality of coverage. The problem is formulated in a Markov Decision Process (MDP) framework and the performance of the transmission policy thus derived is compared with that of an energy balancing policy as well as an aggressive policy. An upper bound on the performance of arbitrary policies is determined and lower bounds specific to the energy balancing and aggressive policies are also developed. Our results show that the quality of coverage associated with the MDP formulation outperforms the other policies.

UNR CSE Host

Dr. Murat Yuksel

Mr. Samir Tamer

Ingersoll Rand Security Technologies

Biography

Samir Tamer researches biometric technologies at Ingersoll Rand Security Technologies, a supplier of biometric and non-biometric security devices. Samir is primarily linked to hand geometry devices used for physical access control and time management. Over 200,000 handreaders have been deployed since the company’s inception in 1986 in such diverse locations as airports, banks, hospitals, and manufacturing floors. Mr. Tamer is a representative to the INCITS M1 Technical Committee on Biometrics and a U.S. delegate to the SC37 ISO/IEC Joint Technical Committee on biometrics. He holds engineering degrees from Duke University and the University of Virginia, and an MBA from Santa Clara University.

Location: MS 321

Friday, March 7th, 2008 at 11:00a.m.

Sponsored and organized by the CSE/EBE/IEEE

Hand Geometry and the Biometrics Industry

When you lock a door, how secure is it really? Keys prove that you HAVE something required to enter. Passwords prove that you KNOW something required to enter. Only biometrics prove that you ARE someone authorized to enter. Biometric devices are those that automatically verify or identify the identity of humans based on biological/behavioral traits such as face geometry, hand geometry, fingerprint patterns, iris patterns, speech patterns, or even vein topologies. This discussion will provide a history of the biometrics market, an introduction to the major biometric modalities mentioned above, and recent trends in the industry (technological advances, corporate consolidation, privacy concerns, and government mandates).

UNR CSE Host

Dr. George Bebis

José P. Zagal

Biography

José Zagal is a PhD candidate in the College of Computing, Georgia Institute of Technology. He has an MSc in engineering sciences and a BS in industrial engineering from Pontificia Universidad Catolica de Chile. His research interests include the use of online communities for collaborative learning and the development of frameworks for describing, analyzing, and understanding games. He is a member of the Electronic Learning Communities Lab and the Experimental Game Lab at Georgia Institute of Technology. In his free time he loves to design and play games.

Location: WRB 2020

Thursday, March 6th, 2008 at 2:00p.m.

Sponsored and organized by the CSE

Exploring the Issues and Challenges of Learning about Videogames

On the surface, it seems like teaching about games should be easy. After all, students are highly motivated, enjoy engaging with course content, and have extensive personal experience with videogames. Games education in reality is surprisingly complex. This talk will discuss the educational and learning issues involved in studying games. It will also describing the design and use of two online learning environments for supporting learning about games: GameLog and the Game Ontology Wiki. GameLog is an online blogging environment designed to help students reflect on their game playing experiences. GameLog differs from traditional blogging environments because each user maintains multiple parallel blogs, with each blog devoted to a single game. The Game Ontology wiki provides an authentic context for students to contribute and participate in a games studies research endeavor: the Game Ontology Project, a hierarchy of elements of gameplay. This talk will address the results of using GameLog and the Game Ontology Wiki in university level games-related classes and discuss the importance that online collaborative environments like these can play in providing students with opportunities for leveraging their personal knowledge of games while helping them achieve a deeper understanding.

Konstantinos Bekris

Location: WRB 3005

Tuesday, March 4th, 2008 at 4:00p.m.

Sponsored and organized by the CSE

Real-Time Kinodynamic Planning: Physically-Realistic, Fast, Safe and Distributed

Many exciting applications, ranging from simulations and games to the control of robots, require real-time computation of collision-free trajectories given only partial knowledge of a potentially dynamic environment. One promising and general approach to address such problems is to employ the sampling-based kinodynamic framework, which can accommodate a variety of systems and directly addresses both geometric and dynamic aspects of motion planning. However, as a search-based approach, it poses computational challenges when time limitations are imposed for real-time performance. This results in low quality paths or partial paths that do not reach the goal. Moreover there are safety considerations, in terms of collision-avoidance, when a system has to respect dynamic motion constraints and operate under time limitations.

This talk describes four contributions in the context of real-time sampling-based kinodynamic planning. Firstly, we incorporate physical simulators in planning so as to be able to better represent realistic dynamics of the physical world, such as drift, friction, contacts and gravity. Secondly, we work on "informed" versions of sampling-based kinodynamic planners, where any vailable workspace information is utilized to reduce solution time, improve path quality and provide a high level guidance in the case of replanning. Thirdly, we propose a "continuous" replanning approach that guarantees the safety of a system with dynamics in tasks with static obstacles. Finally, we have extended this solution to a distributed algorithm for multiple communicating vehicles operating in the same environment. We show that through coordination multiple vehicles can also achieve collision avoidance in real-time while employing sampling-based kinodynamic planners.

Dr. Samee Khan

Biography

Samee U. Khan is a postdoctoral research fellow in the Electrical and Computer Engineering Department at the Colorado State University. He received his Ph.D. degree in computer science from the University of Texas, Arlington in 2007. His research interest include designing, building, analyzing, and measuring large-scale autonomous distributed computing systems using game theoretical and algorithmic mechanism design techniques, passive optical network layouts, designing secure systems, combinatorial games, and combinatorial optimization. His research work in these areas is published in 40 technical papers. Dr. Khan is a member of the European Association of Theoretical Computer Science, the Game Theory Society, the IEEE Communications Society, the IEEE Computer Society, and the Society of Photo-Optical Instrumentation Engineers.

Location: WRB 2025

Monday, February 25th, 2008 at 2:00p.m.

Sponsored and organized by the CSE

Optimizing the Energy Consumption and Performance of Computational Grids

Energy consumption is a critical and crucial problem in large-scale computing systems, such as computational grids because they consume massive amounts of energy and have high cooling costs. These systems must be designed to meet functional and timing requirement while being energyefficient. Resource allocation in computational grids is already a challenging problem due to the need to address deadline constraints and system heterogeneity. The problem becomes more challenging when energy management is an additional design objective because energy consumption of the system must be carefully balanced against other performance measures. This talk focuses on the topic of resource allocation in computational grids with the aim to minimize energy consumption and makespan subject to the constraints of deadlines and architectural requirements. A solution from cooperative game theory based on the concept of Nash Bargaining Solution will be presented. In this game theoretical technique, machines collectively arrive at a decision that describes the best task allocation for the entire computational grid. This collective decision ensures that the allocations are both energy and makespan optimized. We also will look at some experimental results which verify that the cooperative game theoretical technique achieves superior performance compared to other traditional resource allocation techniques.

Dr. Gordon K. Lee

San Diego State University

Biography

Dr. Lee is currently a full Professor in the Department of Electrical and Computer Engineering. His research interests are in the areas of robotics and control systems, particularly intelligent evolutionary control algorithms, fuzzy systems and neural networks, as well as in the applications of these methods to mobile robotic colonies. He has published over 250 technical documents; Dr. Lee is a senior member of IEEE, a member of AIAA and a senior member of ISCA. He is also currently an Associate Editor for the International Journal on Intelligent Automation and Soft Computing.

Location: MS 227

Friday, February 22nd, 2008 at Noon

Sponsored and organized by the CSE/EBE/IEEE

Evolutionary Learning Algorithms for Intelligent Systems

There exist many applications in military, commercial and civilian scenarios where intelligent systems must perform complex tasks in an uncertain environment. For example, autonomous intelligent robot colonies may be used in reconnaissance missions or seek-and-capture scenarios involving a complex set of interactions between machines as well as between machines and humans and may cover long distances to remote sites. Because of the nature of the tasks, new classes of intelligent systems will be required that have a high level of specification for efficiency and reliability. This, we believe, can only be accomplished through sophisticated control and efficient sensor integration as an integral part of the design of the robot and the robot's supporting systems. Evolutionary techniques present a feasible approach to this task. This seminar presents an overview of on-going research in the development of evolutionary learning algorithms that may be applied to a multitude of applications. The talk will include a discussion on adaptive fitness functions, learning algorithms and convergence properties of evolutionary methods.

UNR CSE Host

Dr. Sergiu Dascalu

Dr. Jim Kenyon

Department of Physiology and Cell Biology, UNR

Biography

Dr. Kenyon received his Ph.D. in Physiology and Biophysics in 1977 from the University of Vermont where he identified a component of K+ current important in cardiac excitation. He did post-doctoral work at the University of Maryland studying the details of ion channel gating and then took a faculty position at the University of Texas Health Sciences Center at Dallas where his research focused on cardiac excitation and the activation of contraction. He joined the Department of Physiology and Cell Biology in the University of Nevada School of Medicine in 1987. At Reno, he began studies of the control of intracellular Ca2+ in primary afferent nociceptive neurons, i.e. the neurons where pain starts. He is currently the Director of the Nevada IDeA Network of Biomedical Research Excellence (INBRE), an NIH funded project to develop research infrastructure and personnel in Nevada.

Location: SEM 261

Friday, February 15th, 2008 at 2:30p.m.

Sponsored and organized by the CSE/EBE/IEEE

Information transfer in electrically excitable cells: studies with electrodes, fluorescence, and modeling

Although the electrical nature of signaling in the nervous system was appreciated in the 18th century, essentially no progress toward a useful understanding of the mechanism of this signaling was made until the coming together of a unique experimental preparation, appropriate electrical amplifiers, and prepared minds at the end of the second World War. This presentation will review that event with particular attention to the mathematical model that reconstituted the action potential of the squid giant axon from measurements of voltage-dependent conductances. This conceptual approach led the way for modeling numerous cellular activities with particular success in studies of the control of intracellular calcium. This presentation will review recent data and modeling with particular attention to the current failure of the modeling.

UNR EBE Host

Dr. Sami Fadali

Dr. Tom Malzbender

Hewlett-Packard Laboratories

Biography

Tom is a Senior Research Scientist in the Mobile and Media Systems Lab within Hewlett-Packard Laboratories. Tom works at the intersection of computer graphics, computer vision and signal processing. He has developed the techniques of Reflectance Transformation, Polynomial Texture Mapping and Fourier Volume Rendering. Tom also developed the capacitive sensing technology that allowed HP to penetrate the consumer graphics tablet market. His PTM methods are used by the National Gallery in London, the Tate Gallery and in the fields of criminal forensics, paleontology and archeology. His recent work on imaging the Antikythera Mechanism led to the deciphering of this ancient astronomical computer.

Tom is on the program committee for several 3D graphics and vision conferences. More information can be found at http://www.hpl.hp.com/personal/Tom_Malzbender/ .

Location: OSN 203

Friday, February 1st, 2008 at 11:00a.m.

Sponsored and organized by the CSE/EE/IEEE

Reflectance Imaging: A Simple Approach to Seeing Surface Detail

The appearance of a 3D object depends on viewpoint, lighting conditions, surface shape and reflectance properties. Photography captures the appearance of an object directly, holding the viewpoint and lighting conditions constant. Although simple to capture, photographs sample a limited subset of this "appearance space" and are poorly suited to quantitative analysis. 3D geometry capture of shape is an increasingly popular approach to permit further quantitative analysis of target objects, but is expensive and problematic. Additionally, 3D capture mechanisms are often finicky and fail for locations on the object that are dark or shiny. We present a photographic method, reflectance imaging, that avoids these problems by directly capturing the appearance of an object across lighting variations. Comparable to conventional photography in its simplicity and robustness, it allows quantitative analysis as well as interactive control of lighting, unlike a conventional 2D image. Capturing the reflectance properties of an object also allows enhancements that often allow one to see surface detail not apparent even when directly inspecting the object. We will present 6 years of practical experience with this method in the areas of paleontology, archeology, criminal forensics, artifact and art conservation. Demos and tools useful for experimentation will be described and can be accessed online at www.hpl.hp.com/ptm.

UNR CSE Host

Dr. George Bebis

Mr. Jim Hunt

Bally Technologies

Biography

Jim Hunt has been a quality assurance engineer at Bally Technologies Since July 2002. He manages the companies automated testing groups. Previously, Mr. Hunt was a software consultant specializing in software configuration management, where he analyzed the software development processes of over 200 different companies. Mr. Hunt received a Bachelor of Science degree in electrical engineering from the University of Missouri - Rolla in 1986. Then he gained experience in the design of computer controlled vending and slot machines.

Location: SEM 261

Monday, December 3rd, 2007 at 4:00p.m.

Sponsored and organized by the CSE/EE/IEEE

Essentials of Software Configuration Management

One of the most challenging aspects of software product design is the organization of the artifacts of that design. Configurable items result in each area of the software development life cycle and must be managed across the organization. A brief taxonomy of SCM methods will be presented.

UNR CSE Host

Dr. Sergiu Dascalu

Dr. Philippe Dugerdil

University of Applied Sciences, Geneva, Switzerland

Biography

Since November 2002 Dr. Philippe Dugerdil has been a Professor of Software Engineering in the School of Business of the University of Applied Sciences in Geneva, Switzerland. Between 1993 and 2002 he served as Vice President of Pictet Bank, Geneva, where he headed the front office software development and software engineering departments. Before that, he held positions as group leader and development manager in the software industry and, in late 1980s, worked as an Assistant Professor with the University of Neuchatel, Switzerland. Dr. Dugerdil received an Electrical Engineer degree from the Swiss Institute of Technology, Lausanne, Switzerland (1982), an MBA degree from the Institute for Management Development (IMD), Lausanne (2001), and a PhD degree in Computer Science from the Aix-Marseille University, France (1988). His main research interests are on software engineering and reengineering, program understanding, reverse engineering, and software development processes.

Location: SEM 261

Wednesday, November 21st, 2007 at 4:00p.m.

Sponsored and organized by the CSE/EE/IEEE

Using Execution Trace Segmentation Techniques to Identify Dynamic Clusters of Classes in Legacy Systems

In dynamic analysis (i.e. execution trace analysis), an important problem is to cope with the volume of data to process. However, in the literature, no definitive solution has yet been proposed. Generally, the techniques start by compressing the execution trace before proceeding with the analysis. On the contrary, we propose a way to process the uncompressed execution trace using a segmentation technique. This has the strong advantage to keep the temporal relationships between the calls. First, we present the concept of temporally omnipresent class that is the analogy of the "noise" in signal processing. Then, during analysis, the omnipresent classes can be filtered out to concentrate only on relevant ones. Next, we show a way to recover the components of a legacy system by segmenting the trace. We finally present an application of this approach to a medium size industrial software system, as well as the tool that supports our approach. As a conclusion, we suggest that our noise reduction and clustering techniques are both efficient and scalable.

UNR CSE Host

Dr Sergiu Dascalu

Dr. Andre Bernat

Executive Director, Computer Research Association

Biography

Dr. Andrew Bernat is the current Executive Director of the Computing Research Association (www.cra.org), an association of more than 250 North American academic departments of computer science, computer engineering, and related fields; laboratories and centers in industry, government, and academia engaging in basic computing research; and affiliated professional societies. As founding member and Chair of the Computer Science Department at the University of Texas at El Paso, he developed an acclaimed model of student involvement in research, secured external funding, attracted and hired high quality faculty, and directed the renovation of a building to house the department. In recognition of "... his success in creating arguably the strongest computer science department at a minority-serving institution", the Computing Research Association honored him with the A. Nico Habermann Award. In developing and leading the National Science Foundation-funded Model Institutions for Excellence project at UTEP, he forged working groups across different departments and colleges that dramatically transformed the campus and led to qualitative and quantitative improvement in student achievement. He has led national efforts to increase the participation of underrepresented minorities and women in the computing profession. The workshop series he initiated with colleagues in Mexico dramatically increased the activity and productivity of the Mexican computer science community. While a Program Director at the National Science Foundation, he oversaw the Federal Cyber Service: Scholarship for Service program. His experience ranges from scientific research, with some 62 invited presentations and publications, to educational reform and innovation, with some thirty invited presentations and publications.

Location: HREL 109

Thursday, November 15th, 2007 at 10:30a.m.

Sponsored and organized by the CSE/EE/IEEE

Federal Funding for Computing Research, Building Community and Words from My Sponsor

The Computing Research Association is the Washington-based non-profit looking out for the health of the computing research enterprise by focusing on the people and money necessary to conduct computing research. Since CRA pays my salary, I start with what CRA is doing in general to support the people doing computing research and how students and faculty can make use of these efforts. I then bring people up to date on what is going on with research funding at the federal level and what CRA is doing to improve the situation. This leads into a discussion of the new Computing Community Consortium effort.

UNR CSE Host

Dr. Yaakov Varol

Dr. Felix Schuermann

Biography

Felix Schürmann is the General Project Manager of the Blue Brain Project at the Brain Mind Institute at the Ecole Polytechnique Fédérale de Lausanne(EPFL) in Switzerland. He started his studies of physics at the University of Heidelberg, Germany supported by the German National Academic Foundation. He obtained his master's degree (M.S.) in physics from the State University of New York, Buffalo, USA, under the supervision of Richard Gonsalves. During this time he was a Fulbright Scholar. His master thesis dealt with the foundations of computing: the simulation of quantum computing. He received his Ph.D. in physics from the University of Heidelberg, Germany under the supervision of Karlheinz Meier in 2005. The focus of his work was on alternative approaches to computing: Using mixed-signal VLSI he co-designed an efficient implementation of a neural network in hardware and was the first to adopt "Liquid Computing" in hardware.

Location: MS 321

Friday, November 9th, 2007 at 11:00a.m.

Sponsored and organized by the CSE/EE/IEEE

The Blue Brain Project - Simulation-based Research in Neuroscience

The initial phase of the Blue Brain Project aims to reconstruct the detailed cellular structure and function of the neocortical column (NCC) of the young rat. As a collaboration between the Brain Mind Institute of the Ecole Polytechnique Federale de Lausanne (EPFL) and IBM the project is based on the many years of experimental data from an electrophysiology lab and a dedicated massively parallel computing resource (4-rack BlueGene/L). Over the last 2 years an interdisciplinary team of 35 researchers has cast the reverse-engineering of the biological pieces and the forward construction of detailed mathematical models in an iterative process that allows continuous refinement. The refinement is directed by a bottom-up calibration that aligns the model across all levels - from the ion channels to the emergent network phenomena - with the experimental data. In order to put the expert in the loop, extensive use of visualization and interactive analysis is made, which is powered by a second dedicated supercomputer (SGI Prism Extreme, 16 graphic pipes, 300GB shared memory) in order to realize short turn-around times.

UNR CSE Host

Dr. Fred Harris

Dr. Norimichi Tsumura

Chiba University, Japan

Location: EJCH 205

Friday, November 2nd, 2007 at 11:00a.m.

Sponsored and organized by the UNR Psychology and CSE/EE/IEEE

Appearance Reproduction for Industrial Applications

In the process of product development, the appearance of a product is usually evaluated by directly observing trial pieces. The shape of products can be evaluated by making a mock up or showing a computer graphics image. However, it is difficult to evaluate the actual appearance without making trial pieces, since they are dependent on the viewing devices and environmental illuminant. The evaluation of appearance thus becomes a bottle neck in the production cycle, suggesting the need for better procedures to predict the appearances of products in various industries. In this talk, I will introduce our practical approaches for appearance reproduction in 3D soft proofing, skin color reproduction and e-commerce (Norimichi T., Appearance reproduction and multi-spectral imaging, Color Research and Application Vol. 31, No. 4 pp. 270-277, 2006).

UNR Host

Dr. Michael Webster

Dr. Rafael Fierro

University of New Mexico

Biography

Rafael Fierro received a M.Sc. degree in control engineering from the University of Bradford, UK and a Ph.D. degree in electrical engineering from the University of Texas at Arlington in 1990 and 1997, respectively. He held a postdoctoral appointment with the GRASP Lab, University of Pennsylvania, and a faculty position with the Department of Electrical and Computer Engineering, Oklahoma State University. Dr. Fierro is currently an associate professor in the Electrical and Computer Engineering Department, University of New Mexico. His research interests include hierarchical hybrid and embedded systems, optimization-based cooperative control, and robotics. He is the recipient of a Fulbright Scholarship and a 2004 National Science Foundation CAREER Award.

Location: SEM 201

Friday, October 19th, 2007 at 11:00a.m.

Sponsored and organized by the CSE/EE/IEEE

Coordination Algorithms for Cooperative Dynamic Networks

Recent advances in communication, computation, and embedded technologies support the development of cooperative dynamic networks (e.g., teams of robots, UAVs). The development of these systems is motivated by the recognition that, by distributing computer power and other resources, teams of mobile agents can perform many tasks more efficiently and robustly than an individual robot. For example, teams of robots can complete tasks such as multi-point surveillance, distributed localization and mapping, and cooperative transport. However, currently available coordination schemes are still ad-hoc, and have not yet explored the fundamental limits in terms of achievable performance, energy consumption and operational time in dynamic environments.

In this talk, I will present some methodologies and tools that are being developed to facilitate the design of coordination algorithms for cooperative networks of robots. Additionally, I will briefly describe COMET -- a COoperative MultivehiclE Testbed for research in networked embedded systems. Finally, I will provide some preliminary results on a geometric optimization approach to detecting and intercepting dynamic targets using mobile sensors.

UNR EE host

Dr. Sami Fadali

Dr. Michael Mozer

University of Colorado

Biography

Michael Mozer is a Professor in the Department of Computer Science and the Institute of Cognitive Science at the University of Colorado at Boulder, where he joined the faculty in 1988. Dr. Mozer received his Ph.D. in psychology and cognitive science at the University of California at San Diego and was a postdoctoral fellow at the University of Toronto with Geoffrey Hinton. His primary research focus is on computational models in cognitive neuroscience, particular phenomena of visual attention and its pathologies, perceptual learning, and awareness. His interests include application of machine learning techniques to problems in engineering and artificial intelligence. Dr. Mozer has served on technical advisory boards of multiple start-up companies involved in data mining (including Athene, Umbria Communications, and AnswerOn Technologies), and is a co-founder and research scientist at Sensory Inc. which focuses embedded speech technologies for consumer electronics.

Location: WRB 2023

Friday, October 12th, 2007 at 11:00a.m.

Sponsored and organized by the CSE/EE/IEEE

Rational Models of Cognitive Control

Human behavior is remarkably flexible. An individual who drives the same route each day easily adjusts for a traffic jam or to pick up groceries. Any theory of human cognition must explain not only routine behavior, but how behavior is flexibly modulated by the current environment and goals. In this talk, we discuss this ability, often referred to as cognitive control. We focus on rational models, which argue that cognitive control optimizes performance to the statistical structure of the environment, subject to limitations on current knowledge or processing hardware. We describe how characteristics of the environment and task domain can be estimated from experience, and how these characteristics can then be exploited to make behavior more efficient. We validate our theories via simulation studies that model human data from tasks involving visual search (locating a visual target in a cluttered display) and object categorization.

UNR CSE Host

Dr. Bobby D. Bryant

Carl Franklin, J.D.

Political Science and Criminal Justice, Southern Utah University

Biography

Dr. Franklin is the author of ''Investigator's Guide to Computer Crime'', and several other books on legal and political issues, along with journal articles, including one related to cyber-stalking and another related to municipal response to computer crime. He has worked in the law and the legal profession starting as a Police Officer and then as a Law Clerk, a Prosecutor, and an Attorney. He has been involved in higher education since 1992. More information on his work and experience are available on his website at http://www.suu.edu/faculty/franklinc/

Location: WRB 2030

Wednesday, October 10th, 2007 at Noon

The Computer Crime Investigation

This presentation is on applying Computer Science and the Law to solving various cases and challenges. Included in the presentation are issues related to both hardware and software, identifying a variety of computer crimes and civil issues, describing the criminology related to computer crime, and finally, some examples of both how one can prepare to be a Computer Forensics expert, and the people and the tasks with whom one will work.

Dr. Valerio Pascucci

Lawrence Livermore National Laboratory and U.C. Davis, USA

Biography

Dr. Valerio Pascucci is a Computer Scientist and Project Leader at Lawrence Livermore National Laboratory, Center for Applied Scientific Computing since May 2000 and Adjunct Professor at Computer Science Department of University of California Davis since July 2005. Prior to his CASC tenure, he was a senior research associate at the University of Texas at Austin, Center for Computational Visualization, CS and TICAM Departments. Dr. Pascucci earned a Ph.D. in computer science at Purdue University in May 2000, and a EE Laurea (Master), at the University "La Sapienza" in Roma, Italy, in December 1993, as a member of the Geometric Computing Group. Dr. Valerio Pascucci came to the U.S. in 1995 after having grown up in Roma, Italy.

Location: EJCH 205

Friday, September 28th, 2007 at 11:00a.m.

Sponsored and organized by the CSE/EE/IEEE

Robust Multi-scale Morse Theory for Quantitative Analysis of Massive Scientific Data

Recent advances in scientific computing and data acquisition technologies have produced an "information big bang" that has created a major data analysis and understanding challenge widely acknowledged as a primary bottlenecks in contemporary science. For example, at Lawrence Livermore National Laboratory, which houses several supercomputers including BlueGene/L (the largest supercomputer in the world), scientific simulations routinely generate terabytes of data that must be rigorously explored and analyzed as part of the scientific discovery process. This task cannot be undertaken with classical techniques due to performance barriers. More importantly, current tools lack the robustness necessary to handle the unprecedented complexity of the features represented in these massive datasets.

In this talk I will present recent advances in the use of classical Morse theory for the development of a formally sound and practically robust approach to analyze massive scientific models. In this work we have developed a combinatorial equivalent of smooth topological techniques that retain their formal mathematical foundations while enabling practical implementations that do not introduce numerical approximations. The result is a multi-scale data analysis framework that is provably robust and provides error bounded, quantitative feature extraction and tracking.

I will demonstrate the practical application of this family of techniques for the analysis of Hydrodynamic Instabilities, in which we have identified and quantified for the first time different stages of a turbulent mixing process, and for the analysis of porous media, for which we have provided a new characterization of their structural properties amenable for a multi-scale computational framework.

Presentation Slides

UNR CSE Host

Dr. George Bebis

Dr. Karlene A. Hoo

Texas Tech University, Lubbock, TX

Biography

Dr. Karlene A. Hoo is the Associate Vice President for Research at Texas Tech University, Lubbock, TX and Professor in the Department of Chemical Engineering at the same university. Her research interests encompass modeling of chemical and biochemical processes, dynamics and control, system identification, multivariate statistical analysis, energy efficient plants, energy for sustainability, cardiovascular research as it relates to hemodynamic flow in venous systems, and assistive technologies. She is is the author of 5 invited book chapters, more than 50 journal articles, and more than 100 conference papers. Dr. Hoo advised 6 doctoral and 7 master students who graduated so far and is currently supervising 10 PhD and 7 MS students.

Location: WRB 2003

Friday, March 9th, 2007 at Noon

Sponsored and organized by the IEEE/CHE-MET-EE

Designing Operability into Complex Integrated Processes

The design of a chemical process is a very challenging task due to the high dimensionality, nonlinearity, and the complexities of the material recycle and energy integration. For the process to operate at the designed conditions a control system must itself be designed and integrated with the process. Currently, the process and controller designs are serial tasks. The design of a chemical process is a steady state task with the end product, the process flowsheet, justifed primarily on its economic potential (a steady state balance sheet). There is almost no accounting of controllability, flexibility, and operability - all dynamic propositions - of the designed process even though it is widely known that the process design constrains the achievable control performance.

Modern control systems themselves involves high performance sensing and actuation components that must act synergistically to achieve the demanding tasks of regulation, safety, profit, and effciency. What is not known well and not quantified at all is the dependence that the control system imposes on the achievable performance of the process. Without a priori knowledge and analysis of this bi-directional dependence, the combined pro- cess and control designs may never operate satisfactorily at the designed operating condition.

The problem of designing operability and ultimately controllability into complex integrated processes will be addressed using a particular flowsheet decomposition and plantwide control design method. Appropriate examples will be presented to demonstrate this approach.

UNR CHE-MET Host

Dr. Nicholas Tsoulfandis

Ken Sheppard

PC-Doctor, Inc.

Biography

Ken Sheppard is responsible for all engineering and product development activities at PC-Doctor, Inc. He led the team that built PC-Doctor for Windows, which ships on millions of personal computers made by IBM, Lenovo, HP, Gateway and other leading manufacturers.

Sheppard joined the company as a software engineer in 2000, and quickly rose to senior design engineer. He was appointed to his current position as chief technology officer in May 2005.

Before joining PC-Doctor, Sheppard was in the U.S. Army where he rapidly achieved the rank of captain. He served as information technology and communications officer for a 450-soldier command at Fort Sill, Okla. where he managed a team of 16 soldiers who engineered, installed and maintained an information-technology infrastructure that included more than 200 workstations and six servers. Previously, as a top-rated lieutenant, he conducted fire direction for a howitzer platoon and was responsible for the welfare and training of 40 soldiers.

Sheppard also served as a software engineer for the Hydrologic Engineering Center in Davis, Calif. He is a Sun Certified Programmer for Java 2 and a Microsoft Certified Systems Engineer.

A distinguished graduate of the Reserve Officer Training Corps at the University of California, Davis, Sheppard received a bachelor's degree in computer science and engineering from the university in 1996.

Location: PE 208

Tuesday, March 6th, 2007 at 09:30a.m.

Sponsored and organized by the CSE/EE/IEEE

Pragmatic Development at PC-Doctor

As a software product grows in complexity, maintaining the flexibility to meet fast changing customer requirements and to easily adapt to changes in the industry can become a problem. At PC-Doctor, Inc., a relatively small team of developers needs to maintain, improve, and constantly change a product that needs to support Windows, DOS and UNIX operating systems. The software is also localized in 12 different languages and shipped on millions of PC's each month.

As a way to address this problem, PC Doctor, Inc. uses agile development methodologies, which are designed to facilitate communication and power individuals to make decisions. A subset of these lightweight methodologies that we have distilled down is what we call pragmatic development; these are a set of rules that developers need to constantly keep in mind to ensure they are producing high-quality software as efficiently as possible. In this talk, the set of guidelines will be presented and put into perspective within the domain of everyday use.

UNR CSE Host

Dr. Sergiu Dascalu

Dr. Justin Schonfeld

University of Nevada Reno

Biography

Justin Schonfeld graduated from Oregon State University with Honors Bachelors degrees in Computer Science and General Science in 2000 and from Iowa State University with a Ph.D. in Bioinformatics in 2006. He is currently working as a Postdoc in the Evolutionary and Computation Systems Lab at the University of Nevada, Reno. His research interests include: evolutionary computation, bioinformatics, computational intelligence applied to data mining, and evolved life.

Location: SEM 234

Friday, December 1st, 2006 at Noon

Sponsored and organized by the CSE/EE/IEEE

A modular data analysis pipeline for the discovery of novel RNA motifs

This talk introduces a modular software pipeline that searches collections of RNA sequences for novel RNA motifs. In this case the motifs incorporate elements of primary and secondary structure. The motif search pipeline breaks up sets of RNA sequences into shortened segments of RNA primary sequence. The shortened segments are then folded to obtain low energy secondary structures. The distance estimation module of the pipeline then calculates distances between the folded bricks, and then analyzes the resulting distance matrices for patterns.

An initial implementation of the pipeline is applied to synthetic and biological data sets. This implementation introduces a new distance measure for comparing RNA sequences based on structural annotation of the folded sequence as well as a new data analysis technique called non-linear projection. The modular nature of the pipeline is then used to explore the relationships between several different distance measures on random data, synthetic data, and a biological data set consisting of iron response elements. It is shown that the different distance measures capture different relationships between the RNA sequences. The non-linear projection algorithm is used to produce 2-dimensional projections of the distance matrices which are examined via inspection and k-means multiclustering. The pipeline is able to successfully cluster synthetic RNA sequences based only on primary sequence data as well as the iron response elements data set.

CSE UNR Host

Dr. Sushil Louis.

Joachim Holtz, IEEE Distinguished Speaker

University of Wuppertal

Biography

Joachim Holtz graduated in 1967 and received the Ph.D. degree in 1969 from the Technical University Braunschweig, Germany. In 1969 he became Associate Professor and, in 1971, Full Professor and Head of the Control Engineering Laboratory, Indian Institute of Technology in Madras, India. He joined the Siemens Research Laboratories in Erlangen, Germany, in 1972. From 1976 to 1998, he was Professor and Head of the Electrical Machines and Drives Laboratory, Wuppertal University, Germany. He is presently Professor Emeritus and a Consultant.

Dr. Holtz has published extensively, including 12 invited papers in journals. He has earned 12 Prize Paper Awards. He is the coauthor of four books, and holds 31 patents. He is the recipient of the IEEE Industrial Electronics Society Dr. Eugene Mittelmann Achievement Award, the IEEE Industrial Applications Society Outstanding Achievement Award, the IEEE Power Electronics Society William E. Newell Field Award, the IEEE Third Millenium Medal, and the IEEE Lamme Gold Medal. He is a Fellow of the IEEE.

Dr. Holtz is Past Editor-in-Chief of the IEEE Transactions on Industrial Electronics, Distinguished Speaker of the IEEE Industrial Applications Society and IEEE Industrial Electronics Society.

Location: SEM 331

Monday, November 27th, 2006 at 4:00p.m.

Sponsored and organized by the IEEE

Insightful Dynamic Analysis of AC Machines

AC motors have proliferated as the most important machine type used in speed variable drive systems. The dynamic analysis and description of revolving field machines is supported by well-established theories: Park's transformation (1929), and the space vector theory by Kovacs and Racz (1959). Yet some inconsistencies with the theory of dynamic systems exist: The machine eigenvalues suggest the existence of two damped oscillators; it appears unsatisfactory that the respective eigenfrequencies change with the velocity of the reference frame. This contradicts the common understanding according to which the eigenfrequency is an inherent system property.

A clarification is reached using a novel approach for the dynamic analysis. The approach is based on complex state variables. It permits relating a transient condition to the propagation processes in space of distributed magnetic fields. The formal analysis constitutes an extension to the space vector theory and to the theory of dynamic systems. Its application eases the design of closed loop control systems for ac machine drives.

UNR EE Host

Dr. Andy Trzynadlowski

Dr. Vic Grout and Mr. Stuart Cunningham

Centre for Applied Internet Research, University of Wales, UK

Biography

Dr. Vic Grout

Vic Grout was awarded the BSc(Hons) degree in Mathematics and Computing from the University of Exeter (UK) in 1984 and the PhD degree in Communication Engineering from Plymouth Polytechnic (UK) in 1988.

He has worked in senior positions in both academia and industry for twenty years and has published and presented over 100 research papers. He is currently a Reader in Computer Science at the University of Wales NEWI, Wrexham in the UK, where he leads the Centre for Applied Internet Research (CAIR). His research interests and those of his research students span several areas of computational mathematics, particularly the application of heuristic principles to large-scale problems in network design and management.

Dr. Grout is a Chartered Engineer, Chartered Scientist and Chartered Mathematician, a member of the IMA, IEE, ACM and IEEE and a Fellow of the British Computer Society (BCS). He chairs the biennial international conference series on Internet Technologies and Applications (ITA).

Stuart Cunningham

Stuart Cunningham was awarded his BSc degree in Computer Networks in 2001, and in 2003 was awarded the MSc Multimedia Communications degree with Distinction, both from the University of Paisley (UK). He is a Member of the British Computer Society and the Institute of Engineering & Technology. Stuart is also a member of the MPEG Music Notation Standards working group.

Since 2003, he has been working at the University of Wales as a lecturer where he teaches audio visual computing and computer systems architecture. Stuart is also a PhD student at the University of Wales, studying under the supervision of Dr. Vic Grout.

Location: SEM 261

Wednesday, November 15th, 2006 at 4:00p.m.

Sponsored and organized by the CSE/EE/IEEE

Optimising Internet Access Control Lists

Access Control Lists (ACLs) are becoming increasingly widespread in networking, both in their number and range of use. An ACL, or packet filter, consists of an ordered sequence (or some other structure) of rules, each rule being of the form permit A B C ... or deny X Y Z ..., where A B C ... X Y Z are characteristics of the traffic being filtered. Typical characteristics could be the source or destination address (or range of addresses) or packet type (eg, IP, ICMP, TCP, UDP, etc.). More sophisticated characteristics may be used, however, and combinations are also normal. Each incoming (or sometimes outgoing) packet is tested against each rule until a match is found, at which point the packet is permitted or denied accordingly.

The original use of ACLs, as the rule notation suggests, was simply to pass or block traffic, maybe of a specified type, to or from certain parts of an internet. However, ACLs now have a much wider purpose in selecting packets for any traffic policy to be applied at key points in or between domains. Traffic shaping, tunnelling, NAT, policy-based routing, etc. all use ACLs or their equivalents to select packets to which to apply the policy (and to ignore the rest). It is common now for a packet to be matched against several ACLs across a single router or switch, more through a domain and many across an internet.

This all takes time; more time for larger lists and more again for more lists. The increased use of ACLs can increase packet latency across an internet beyond acceptable limits. There is clearly merit in attempting to optimise this process - to find the best structure and combination of ACLs to achieve a specified purpose. Unfortunately, this is no simple objective. There are partial solutions in both hardware and software (and combinations of both), working both on- and off-line, but no utopia as yet. This talk considers the problem from first principles, discusses possible approaches and suggests some directions for the future.

UNR CSE Host

Dr. Sergiu Dascalu

Mr. Jeff Elpern

Reno/Tahoe Software Quality Institute

Biography

Jeff Elpern is a high-tech executive and entrepreneur. Currently he is the V.P. of Software for a Silicon Valley telecommunications component manufacturer, founder and CEO of the Software Quality Institute (SQI, Inc.) based in Reno, and founder of the non-profit Open Source Nevada. He was founder of two startups and on the executive teams for two other start-ups. Earlier in his career, he ran the largest quantitative marketing operation on Madison Avenue and was part of the Lee Iacocca turn-around team at Chrysler. He is a native Nevadan. He did his undergraduate work at UNR and received a Masters of Science in Quantitative Analysis from Carnegie Mellon University.

Location: SEM 261

Monday, October 30th, 2006 at 4:00p.m.

Sponsored and organized by the CSE/EE/IEEE

A Framework for Understanding the Open Revolution

The Open Source Revolution is already having dramatic impact on the computer industry. Web services based on Open Source technologies play a major role in the Internet. The Linux operating system has achieved the dominating position within the embedded controller segment of the telecommunication industry. And, recently, Open Source applications have passed Mac applications in penetration into the PC market. Why is this happening? Should we be surprised? Is it a sustained phenomenon?

I'll present a framework for understanding the Open Source Revolution by identifying a number of market forces driving the revolution and placing these forces within historical perspective. I'll show that the Open Source Revolution is a natural response, and part of a continuing effort, by Users to increase their returns from technology by controlling the market power of commercial software developers. The core argument is based on economics. As Users pursue maximizing the economic returns of their software portfolios, they gravitate toward software solutions that limit the market power of commercial developers. An example of this is the movement toward more and more standards. The adoption of Open Source is a natural next step for Users in the battle for the control of market power.

Thus, the Open Source Revolution is the current "front line" in the battle between software developers and Users on how economic returns from technologies are allocated between the two. In addition, Open Source will be shown to be a "Disruptive" technology - as defined by Clayton Christensen in "The Innovator's Dilemma." This market force explains the "why now" issue. As the current commercial software leaders' efforts for "Sustaining Technology Innovations" exceeds the User's ability to absorb new features and power, the seeds for the entry of a disruptive technology are sowed. Open Source fits all three criteria for a disruptive technology, which will be discussed in the presentation. It is also important to note that a paradigm shift like this, the shift from proprietary code to Open Source, always changes the face of winners and losers, and this will affect everyone in the industry.

Finally, my presentation will include scenarios of why people and corporations participate in Open Source by defining Open Source business models, current and emerging.

UNR CSE Host

Dr. Sergiu Dascalu

Dr. Risto Miikkulainen

University of Texas at Austin

Biography

Risto Miikkulainen is a Professor of Computer Sciences at the University of Texas at Austin. His recent research focuses on methods for evolving neural networks and applying these methods to game playing, robotics, and intelligent control. He is an author of over 200 articles on neuroevolution, connectionist natural language processing, and the computational neuroscience of the visual cortex. He is an editor of the Machine Learning Journal and Journal of Cognitive Systems Research.

Location: REL 109

Friday, October 13th, 2006 at Noon

Sponsored and organized by the CSE/EE/IEEE

Solving Sequential Decision Tasks With Neuroevolution

Neuroevolution, or training neural networks through genetic algorithms, is a policy-iteration method that can potentially solve difficult reinforcement learning tasks. Recurrent neural networks can be evolved to map sequences of states directly to optimal actions, which is a robust approach with continuous domains and with hidden states. In this talk, I will review recent advances in neuroevolution methods, and present several applications ranging from rocket control and autonomous vehicles to robotics and games.

UNR CSE Host

Dr. Bobby Bryant

Dr. Kenneth O. Stanley

University of Central Florida

Biography

Kenneth O. Stanley is an Assistant Professor in the School of Computer Science at the University of Central Florida. His research focuses on artificially evolving complex solutions to difficult real-world tasks. He graduated magna cum laude with a B.S.E. in Computer Science Engineering and a minor in Cognitive Science from the University of Pennsylvania in 1997. He received an M.S. in Computer Science in 1999, and a Ph.D. in 2004 at the University of Texas at Austin. He has won best paper awards for his work on NEAT (at the 2002 Genetic and Evolutionary Computation Conference) and for his work on NERO (at the IEEE 2005 Symposium on Computational Intelligence and Games), and also won the Independent Games Festival Student Showcase Award (at the 2006 Game Developers conference) for NERO. He has published papers in JAIR, Evolutionary Computation, IEEE Transactions on Evolutionary Computation, and Artificial Life journals.

Location: REL 109

Friday, October 6th, 2006 at Noon

Sponsored and organized by the CSE/EE/IEEE

Real-time Neuroevolution and the NERO Video Game

A major goal for AI is to allow users to interact with agents that learn in real time, making possible new kinds of interactive simulations, training applications, and digital entertainment. This talk will introduce such a learning technology, called real-time NeuroEvolution of Augmenting Topologies (rtNEAT), and describes how rtNEAT was used to build a new genre of video game in which the player teaches agents in real time to perform novel tasks. The game, NeuroEvolving Robotic Operatives (NERO), took a team of over 30 volunteer programmers and artists over two years to create and has received significant recognition since its release. This talk explains the rtNEAT method for evolving increasingly complex neural networks and the NERO video game. Providing laymen the capability to effectively train agents in real time with no prior knowledge of AI or machine learning has broad implications, both in promoting the field of AI and making its achievements accessible to the public at large.

UNR CSE Host

Dr. Bobby Bryant

Dr. Bill Labate

Associate Director for Research Computing Technologies, University of California, Los Angeles

Biography

Dr. Bill Labate is the Associate Director for Research Computing Technologies with UCLA's Office Of Information Technology. He is also the Vice-Chairman of the University of California Research Computing Group and has worked in the information technology field for over twenty-five years with technical and management positions in the private sector, defense industry, private consulting as well as academia. Bill is also the project manager for the UC Grid as well as the UCLA Grid Project.

Location: SEM-234

Friday, September 29th, 2006 at Noon

Sponsored and organized by the CSE/INBRE/IEEE

The UC Grid Portal: Locally Managed, Virtually Available

The University of California system is working to build out a UC Grid Cyberinfrastructure. The goal of the project is to enable access to a vast array of high performance computing, visualization, and storage resources located throughout the UC system regardless of location and to make cross-campus and cross-disciplinary teams from an HPC standpoint a reality. The project sets the course for jointly aggregating and building resources to achieve greater capabilities for individual researchers. The ultimate vision is an overlay to HPC resources world-wide to enable secure, unified, "anytime/anywhere" access to resources formerly available only to researchers who could master the complexities of running HPC systems and had the budgets and manpower to maintain them. This talk will focus on an overview of the challenges the UC system is currently facing, what we hope to achieve with the Grid, a discussion of the architecture of both the campus and UC Grids, the resource "pool" concept, current capabilities and future enhancements, and a hardware and manpower requirement breakdown.

UNR CSE Host

Dr. Fred Harris

Dr. Murray Campbell

Research Scientist, IBM T.J. Watson Research Center, Yorktown Heights, NY

Biography

Dr. Campbell is a research scientist at the IBM T.J. Watson Research Center in Yorktown Heights, New York. He is a member of the team that developed Deep Blue, the first computer to defeat the World Chess Champion in a regulation match, for which he was awarded the Fredkin prize and the Allen Newell Research Excellence Medal. Dr. Campbell received his Ph.D. in Computer Science from Carnegie Mellon University in 1987. He was a strong expert-level player while still in high school and brought this knowledge to the development of Deep Blue's evaluation function--the component of Deep Blue that assesses the value of the current position in a game. He also worked closely with the team's chess consultant, international grandmaster Joel Benjamin, in developing Deep Blue's opening book. Dr. Campbell is now manager of the Intelligent Information Analysis group at IBM, which focuses on analysis of real-time sensor data for early warning applications, as well as indexing and searching of multimedia data.

Location: IGT Training Center, The Wisdom World room

Tuesday, May 23rd, 2006 at 1:30p.m.

Sponsored and organized by the 2006 IGT-UNR Symposium

Looking Back at Deep Blue

It has been nine years since IBM Research's Deep Blue defeated Garry Kasparov, the thenreigning world chess champion, in an epic six-game match that was closely watched by millions. In this talk I will present the background that led up to the decisive match, review the match itself, and discuss some of the broader implications of Deep Blue's victory. Issues I will cover include Deep Blue's connections to high-performance computing, what “intelligence” really means, and the roles that games play in the fields of artificial intelligence, education, and entertainment.

UNR Event

IEEE CIG'06 Symposium

Dr. Norm Brown

CEO, Quadrant-One, Inc., Washington, DC

Biography

Dr. Brown began life as a software developer and manager, then went on to tackle the $42 Billion spent annually by the Department of Defense on developing software for its weapons systems and other needs. He founded the Department of Defense's Software Program Managers Network, and its Airlie Software Council, recruiting the likes of Tom Demarco, Ed Yourdan, Tom McCabe, Capers Jones, and others to reformulate how to improve development of largescale software. Dr. Brown advised the Undersecretary of Defense and served with the Defense Science Board Software Study. What they found is, not surprisingly, directly applicable to virtually all large-scale software development, including gaming, and key implications of these findings will be revealed in his presentation.

Location: IGT Training Center, !e Wisdom World room

Tuesday, May 23rd, 2006 at 11:00a.m.

Sponsored and organized by the 2006 IGT-UNR Symposium

How Dark Matter Affects Software Development, and How to Deal with It

Software Dark Matter typically comprises a significant portion of software development efforts. Such Dark Matter consumes developer time and effort, along with test resources; and, as with ordinary galactic Dark Matter, is typically invisible — usually only discernable by its effect upon developers and its substantial contribution to unnecessary development delays and additional unnecessary efforts. Two high-leverage techniques for identifying and reducing the amount of Software Dark Matter in your development and consequently creating a happier workplace, improving development schedules, and reducing cost, will be addressed.

UNR Event

IEEE CIG'06 Symposium

Dr. Henry Markram

EPFL, Lausanne, Switzerland

Biography

Dr. Henry Markram moved to EPFL in 2002 as full professor. From 1995 to 2001, he was at the Weizmann Institute where he received early tenure and was Stanley and Hellen Diller Professor of Neuroscience. In 1994-95, he was Minerva Fellow in Laboratory of Nobelist Bert Sakmann at the Max Planck Institute, where he discovered calcium transients in dendrites evoked by sub-threshold activity, and by single action potentials propagating back into dendrites. He also began studying the connectivity between neurons and published a paper describing in great detail how layer 5 pyramidal neurons are inter-connected. In 1992-93, he was Senior Fulbright Scholar at the National Institutes of Health (NIH), where he studied ion channels on synaptic vesicles.

He was the first to alter the precise millisecond relative timing of single pre and postsynaptic action potentials to reveal a highly precise learning mechanism operating between neurons which has now been reproduced in many brain regions and is now commonly know as spike timing-dependent synaptic plasticity (STDP). At Weizmann, he started systematically dissecting out the neocortical column, and discovered that synaptic learning can also involve a change in synaptic dynamics rather than merely changing the strengths of connections. He also revealed a spectrum of new principles governing neocortical microcircuit structure, function, and emergent dynamics. Based on the emergent dynamics of the neocortical microcircuit he, together with Wolfgang Maass developed the theory of liquid computing or high entropy computing. At the BMI, he has continued to unravel the blue print of the neocortical column at a greatly accelerated pace building the state of art tools to carry out multi-neuron patch clamp recordings combined with laser and electrical stimulation as well as multi-site electrical recording (up to 12 patch-clamp recordings) and chemical imaging and gene expression.

He has received numerous awards, including the Ebner Science Award in 2001, the James Heinemann Prize in 1999, and the Abramson Research Award in 1998, and has published over 75 papers. In April, 2005 the EPFL signed the agreement with IBM to launch one of the largest single initiatives in neuroscience — the Blue Brain Project.

Dr. Henry Markram obtained his B.Sc. (Hons) from the University of Cape Town, South Africa, under the supervision of Rodney Douglas and his Ph.D from the Weizmann Institute of Science, Israel, under the supervision of Menahem Segal.

Location: SEM 261

Friday, May 12th, 2006 at 10:45a.m.

Sponsored and organized by the UNR Brain Computation Lab & the Office of the VP for Research

The Frontier of Cognitive Computing: The Blue Brain Project

The Blue Brain Project was launched to make the first step towards building a biologically accurate software model of the brain of mammals, eventually including that of man. The seminar will describe the first phase of the project which is to build a neocortical column consisting of 10,000 morphologically complex neurons interconnected with around 20 million synapses each precisely placed on neurons in 3D space.

The experimental basis for the model as well as the technology platform that uses Linux machines, PCs, and two different supercomputers; IBM's Blue Gene and SGI's Altix Extreme Series will be described. Results from the first simulations of the Blue Column will also be presented.

These simulations provide the first hints at a revolutionary new theory of how the brain may be building perceptions and how we may need to reinterpret the past 50 years of experimental results.

We also believe that biologically accurate models of this part of the neocortex would provide the key foundation to build software models of mammalian brains, which could open up new ways of exploring brain function and causes of neurological and psychiatric disorders as well as new diagnostic methods.

UNR CSE Host

Dr. Fred Harris

Dr. Salil Prabhakar

Digital Persona, Inc.

Biography

Salil Prabhakar received his B.Tech. degree in Computer Science and Engineering from Institute of Technology, Banaras Hindu University, Varanasi, India, in 1996. During 1996-1997 he worked with IBM India as a software engineer. He received his Ph.D. degree in Computer Science and Engineering from Michigan State University, East Lansing, MI 48824, in 2001. He currently leads the Algorithms Research Group at Digital Persona Inc., Redwood City, CA 94063 where he works on fingerprint based biometric solutions.

Location: AGN (SEM 201)

Friday, May 5th, 2006 at 4:00p.m.

Sponsored and organized by the CSE/EE/IEEE

Fingerprint Recognition: Future Directions

Many researchers in the pattern recognition community perceive automatic fingerprint recognition to be a solved problem due to the fact that first successful system was deployed over 30 years ago. But contrary to this notion, fingerprint recognition remains a very challenging and exciting pattern recognition research problem with many open issues and research opportunities, solving which will have profound security and economic implications. In this talk, I will talk about some of the challenges in fingerprint recognition and some possible future research directions.

UNR CSE Host

Dr. George Bebis

Mr. Tamer Uz

University of Nevada, Reno

Location: AGN (SEM 201)

Friday, May 5th, 2006 at 2:00p.m.

Sponsored and organized by the Department of Computer Science & Engineering

Fingerprint Template Synthesis

Additional event details will be made available shortly

UNR CSE Host:

Dr. George Bebis

Mr. Uday Rajanna

University of Nevada, Reno

Location: AGN (SEM 201)

Friday, May 5th, 2006 at Noon

Sponsored and organized by the Department of Computer Science & Engineering

Improving the Performance of Fingerprint Classification

Additional event details will be made available shortly

UNR CSE Host

Dr. George Bebis

Computer Science & Engineering

University of Nevada, Reno

Location: REL 109 & 110

Friday, May 5th, 2006 at 10:30a.m.

Senior Projects

The CS426 and CPE426 Senior Projects represent innovative software and hardware products designed and developed during the year by undergraduate Computer Science & Engineering students under the supervision of Dr. Sergiu Dascalu. Project topics include a framework for behavior-based robotics control, an X10-based home automation system, a software tool for data visualization, a program for student budgeting, a web-based interactive floor plan mapper, a first-person shooter video game, and a virtual reality system for puppet control and animation.

More information about this event may be found on the Workshop Flyer and the Workshop Schedule.

Bobby Bryant, Ph.D. Candidate

University of Texas

Location: SEM 326

Thursday, May 4th, 2006 at 09:30a.m.

Sponsored and organized by the IGT Distinguished Speakers Series

Intelligent Agents for Games and Simulators: Creating Adaptive Teams with Guided Evolution

Autonomous intelligent agents that operate in visible environments such as video games and simulators must behave in ways that viewers find convincingly intelligent. For most agents that behavior needs to be robust, flexible, disciplined, and self-consistent. In this talk I will propose methods for inducing such properties into the behavior of a team of agents operating in a video game. Training with neuroevolution, i.e. evolving neural networks with genetic algorithms, provides the desired robust and flexible behavior, but human-generated examples are needed to make the agents disciplined and self-consistent. Combining examples with evolution makes it possible to induce visibly intelligent behavior in autonomous agents via machine learning, making gameplay more satisfying and simulation environments more realistic.

UNR Host

Department of Computer Science & Engineering

Mehran Asadi

The University of Texas at Arlington

Biography

Mehran Asadi is a PhD candidate in the department of computer science and engineering at the University of Texas at Arlington (UTA). He received his B.S. degree from Amir Kabir University of Technology, Tehran, Iran, in Applied Mathematics and Computer Science and his M.S. degree from the University of Texas at Arlington, in Computer Science and Engineering. From 1994 to 2001, he was with the Semi Conductor Lab, in Iran Electronic Industries and SAPCO Co., Tehran, Iran. His current research interests are in the areas of Intelligent Systems, Game/AI, and Autonomous Systems.

Mehran has been a research associate in Artificial Intelligence laboratory at UTA since 2001, and member of AAAI and IEEE.

Location: MS 227

Tuesday, May 2nd, 2006 at 2:30p.m.

Sponsored and organized by the IGT Distinguished Speakers Series

Structural Knowledge Transfer in Reinforcement Learning Using Action-Dependent Partitioning

Autonomous systems are often difficult to program. Reinforcement learning (RL) is an attractive alternative, as it allows the agent to learn behavior on the basis of sparse, delayed reward signals provided only when the agent reaches desired goals.

Recent attempts to address the dimensionality of RL have turned to principled ways of exploiting temporal abstraction, where decisions are not required at each step, but rather invoke the execution of temporally-extended activities which follow their own policies until termination. This leads naturally to hierarchical control architectures and associated learning algorithms.

We presents a new method for the autonomous construction of hierarchical action and state representations in reinforcement learning, aimed at accelerating learning and extending the scope of such systems. In this approach, the agent uses information acquired while learning one task to discover subgoals for similar tasks. The agent is able to transfer knowledge to subsequent tasks and to accelerate learning by creating useful new subgoals and by learning of corresponding subtask policies as abstract actions (options). At the same time, the subgoal actions are used to construct a more abstract state representation using action-dependent state space partitioning.

This representation forms a new level in the state space hierarchy and serves as the initial representation for new learning tasks. In order to ensure that tasks are learnable, value functions are built simultaneously at different levels of the hierarchy and inconsistencies are used to identify actions to be used to refine relevant portions of the abstract state space.

Together, these techniques permit the agent to form more abstract action and state representations over time. Experiments in Smart Home environment and Computer Game domains show that the presented method can significantly outperform learning on a flat state space representation.

UNR Host

Department of Computer Science & Engineering

Beifang Yi

University of Nevada, Reno

Location: SEM 326

Friday, April 28th, 2006 at 3:00p.m.

Sponsored and organized by the CSE Seminar

A Framework for a Sign Language Interfacing System

Additional event details will be made available shortly

UNR CSE Host

Dr. Fred Harris

Wenji Mao, Ph.D. Candidate

University of Southern California

Biography

Wenji Mao is a Ph.D. candidate in the Computer Science Department at the University of Southern California. She has been working at the USC Institute for Creative Technologies since fall, 2001. Her research is focused on AI, intelligent agents and cognitive modeling for virtual training and interactive entertainment. Prior to joining USC, she was a researcher at the German Research Center for Artificial Intelligence (DFKI GmbH) and a lecturer at the Graduate School of Chinese Academy of Science, where she obtained her M.Sc. in Computer Science. She received her M. Eng. from the University of Southern California in May, 2003.

Location: MS 321

Friday, April 28th, 2006 at 2:00p.m.

Sponsored and organized by the IGT Distinguished Speaker Series

Modeling Multi-Agent Social Interaction in Virtual Training and Entertainment

With the advance of multi-agent interactive systems, user-centric adaptive interfaces and systems that socially interact with people, it is increasingly important to model rich social interactions among intelligent entities (agents and human). In such context, a central issue is to build realistic agent models that mimic the cognitive process and inference of how people evaluate social events so as to drive believable behavior generation for intelligent agents.

In this talk, I will present my work on developing a domain-independent computational framework of social causality and social judgment in multi-agent interactions, which is part of the ongoing effort of creating human-like virtual characters in interactive training and entertainment at the University of Southern California. Based on attribution theory in social psychology, my model formalizes the commonsense reasoning about the beliefs of attribution variables such as physical cause, intention, foreknowledge and perceived coercion, from the observations of natural language communication and task execution. In addition, I have designed and conducted experiments to empirically validate the model, using real human data. The validated computational model can be generally incorporated into an intelligent interactive system to augment its cognitive and social functionality.

Finally, I will talk about the applications of my work in emotion modeling, natural language conversation strategies and enhancing coherence in virtual humans, as well as some future research directions.

UNR Host

Department of Computer Science & Engineering

Dr. Eelke Folmer

University of Alberta

Biography

Eelke Folmer is a postdoctoral fellow in the Department of Computer Science at the University of Alberta. He received a Ph.D. degree from the University of Groningen where he worked on investigating the relationship between software architecture and usability for the European union funded research project called STATUS (Software Architecture for Usability). His current research activities include software architecture design, software quality and interaction design specifically focusing on the domain of computer games.

Location: REL 110

Thursday, April 27th, 2006 at 10:00a.m.

Sponsored and organized by the IGT Distinguished Speakers Series

Game Engineering

Computer games have grown considerably in scale and complexity since their humble beginnings in the 1960s. Modern day computer games have reached incredible levels of realism, especially in areas like graphics, physical simulation, and artificial intelligence. A natural consequence of this increase in scale is that the resources required to produce games has also significantly increased. Currently, the average cost for developing a console game is estimated to vary between $3 million and $10 million. Costs have become a serious problem for game developers as the games market is hits driven; the top 99 titles (only 3.3% of development) account for 55% of all sales.

Despite significant advances in software engineering, the development of computer games generally does not employ state-of-the-art software engineering practices and tools. Games development is typically characterized by schedule slips and budget overruns. A much heard-complaint within the game industry is that existing software engineering methods are not adapted to the games domain; for example the creative and exploratory nature of developing games favors an iterative methodology, while the business of games favors a waterfall methodology. Mismatches occur when engineering methodologies are applied to engineer elusive game qualities such as fun. Because the game business is risky and a game company is only as successful as its last release; as a result, the importance of short-term goals is exaggerated and investing in SE practices, is simply not cost effective in the short run. However there is definitely an interest in the games industry in investing in better SE practices. Though, as is common throughout the wider IT industry; high ideals are often abandoned under the pressure of deadlines and shipping products, which is higher in the games industry then in any other IT sector.

In my research, I study the current problems in the games industry and develop software engineering methods, tools and techniques adapted to the games domain, which are the key to lowering costs, reducing time to market and increasing the engineering quality of games.

I this talk I will present some of my current multidisciplinary research activities which include reference architectures and interaction design patterns for games. I will also discuss some promising future research directions.

UNR Host

Department of Computer Science & Engineering

Murat Yuksel

Rensselaer Polytechnic Institute, Troy, NY

Biography

Murat Yuksel is currently a Research Assistant Professor at the Electrical, Computer, and Systems Engineering Department of Rensselaer Polytechnic Institute (RPI), Troy, NY, where he was a Postdoctoral Research Associate and Adjunct Faculty in 2002-2005. He received a B.S. degree in Computer Engineering from Ege University, Izmir, Turkey, in 1996. He, then, received M.S. and Ph.D. degrees in Computer Science from RPI in 1999 and 2002 respectively. His research interests lies in the area of computer communication networks with a special emphasis on wireless sensor and ad-hoc networks, large-scale network simulation and experiment design, free-space-optical mobile ad hoc networks, optimization heuristic algorithms, lossless networks, network economics and performance analysis. He is a member of IEEE and Sigma Xi. He has three pending patents, and has published 26 journal and conference papers.

Location: SEM 234

Thursday, March 16th, 2006 at 3:00p.m.

Sponsored and organized by the IGT Distinduished Speaker Series

Routing and Communication Building Blocks for Multi-hop Wireless Ad Hoc Networks

Nodes in ad hoc wireless networks are typically low-memory, low-powered; and they cannot maintain routing tables large enough for well-known proactive routing protocols. Therefore, stateless and greedy forwarding at intermediate nodes is desirable in ad-hoc networks. Also, for end-to-end traffic engineering, multi-path capabilities and flexibility in routing are necessary building blocks. So, it is desirable to define routes at the source like in Source Routing (SR) while performing greedy forwarding at intermediate nodes. To address these issues, we present our research on Trajectory-Based Routing (TBR) which is a middle-ground between SR and greedy forwarding techniques. We address various issues regarding implementation of TBR. In particular, we (1) develop a trajectory encoding technique that uses the well-known Bezier parametric curves, (2) develop an optimal forwarding strategy and show its superior performance in comparison to other greedy forwarding techniques, and (3) propose architecture to implement long and/or complex trajectories that cannot be encoded in a small size packet header.

In addition to routing difficulties caused by lack of resources, communication environment presents significant challenges to high-speed wireless networking. Legacy RF signals provide omni-directional connectivity, but they consume large transmission power and are limited in communication bandwidth. In comparison, free-space-optical (FSO) communication devices have enormous bandwidth, but line-of-sight (LOS) and alignment issues make them hard to deploy. We present inexpensive FSO node designs that solve the LOS and alignment problems for multi-hop wireless communication. We will also show that our FSO designs can significantly enhance the higher-layer networking operations such as routing.

Ann Gordon-Ross

University of California, Riverside

Biography

Ann Gordon-Ross is currently a Ph.D. student in the Department of Computer Science and Engineering at the University of California, Riverside. She received her BS in Computer Science from the University of California, Riverside in 2000. Her research interests include dynamic optimizations for low power embedded and multi-core systems.

Location: SEM 344

Monday, March 13th, 2006 at 3:00p.m.

Sponsored and organized by the IGT Distinguished Speaker Series

Dynamic Optimization of Highly-Configurable Caches for Reduced Energy Consumption

Caches may consume a large percentage of microprocessor system power — as much as 50% in some systems — and thus reducing cache power has been an active research area during the past decade. Because the best cache design may vary tremendously for different applications, configurable caches have recently been proposed and have appeared commercially, especially in embedded systems. However, quickly finding the best cache configuration for a particular application is a challenging problem. This talk describes methods to efficiently find near-optimal cache configurations, for a number of progressively more complex configurable cache hierarchies. The methods not only apply to approaches where optimization is done beforehand using simulations, but they can even be applied dynamically for completely transparent self-tuning cache hierarchies. The methods find near-optimal configurations that yield memory-access energy savings of 65% on average, while executing quickly due to searching a mere 0.002% of the search space.

Greg Stitt

University of California, Riverside

Biography

Greg Stitt's research interests include binary synthesis and hardware/software partitioning, low-power synthesis, low-power architectures and compilers.

Location: SEM 261

Friday, March 10th, 2006 at 3:00p.m.

Sponsored and organized by the IGT Distinguished Speaker Series

Synthesis from Software Binaries

Over the past decade, many high-level synthesis and hardware/software partitioning approaches have been proposed to automatically improve the performance of embedded applications. However, these approaches have yet to achieve wide commercial success, due in part to the difficulty of incorporating such approaches into software tool flows. The requirement of using a specific language, compiler, or development environment may cause many software developers to resist such approaches, due to the difficulty and possible instability of changing well-established robust tool flows. Thus, we introduced synthesis from binaries as a means of better integrating with tool flows, due to binary synthesis can be more easily integrated into a software development tool-flow by only requiring an additional back-end tool. However, binary synthesis potentially is less effective than traditional approaches because of the lack of high-level information in a binary. In this talk, I will discuss the key techniques that make binary synthesis competitive with traditional high-level approaches. In addition I will discuss future research directions that are enabled by binary synthesis, including warp processors, which are microprocessors that are capable of dynamically and transparently optimizing a software application by synthesizing hardware at runtime.

Dr. Marcel Karam

American University of Beirut, Lebanon

Biography

Dr. Marcel Karam received his PhD in 2002 from the Faculty of Computer Science at Dalhousie University. Dr. Karam is currently an Assistant Professor with the Department of Computer Science at the American University of Beirut. He is an active researcher in the testing theories and practices of different programming paradigms, in particular of visual dataflow languages. He is also interested in applying visualization to model view workflow (MVWf) web engineering languages and environments. During his time with IBM Watson, much of his work revolved around the design and implementation of Effigy, a visual object oriented conceptual language. His research in computer science has been supported by several LNRC research grants and has resulted in numerous journal and conference publications. He has served on technical program committees of various software engineering conferences and workshops.

Location: AGN (SEM 201)

Thursday, March 9th, 2006 at 4:00p.m.

Sponsored and organized by the CSE/EE/IEEE

Visual Reflections in Test Coverage

The most important characteristic of visual notations is their capacity for providing a concrete view of a problem, termed “closeness of mapping.” Unlike visual dataflow languages that use iconic representations of the program constructs, with imperative programs locating variable definition-use interactions that are associated with program faults is one of the most difficult and time-consuming tasks of the intra-procedural debugging process. This task is further complicated by the presence of loops and aliases. A new visualization technique will be presented in an effort to assist with testing tasks. The technique employs a mapping scheme that uses visual artifacts (“visual reflections”) and colors to graphically map the participation of each statement, control, and variable DU-chains in the outcomes of procedure's execution.

UNR CSE Host

Dr. Sergiu Dascalu

Dr. Brian W. Beck

University of Nevada, Reno Center for Bioinformatics
Department of Biochemistry & Molecular Biology

Biography

Dr. Brian W. Beck is the Associate Director of Molecular Modeling for the Center for Bioinformatics and the Department of Biochemistry & Molecular Biology at the University of Nevada, Reno. He received a BS in Biochemistry from Texas A&M University and a Ph.D. in Biochemistry & Biophysics from Washington State University where he worked with Prof. Toshiko Ichiye on computational analysis (particularly molecular dynamics simulations) of the manner in which electron transfer proteins regulate their redox potentials. His current work involves the use and development of computational techniques for the analysis of the relationship biomacromolecular structure to function, particularly the identification and analysis of protein:protein interfaces and the effects of solvent on protein structure.

Location: SEM 201

Thursday, February 23rd, 2006 at 4:00p.m.

Sponsored and organized by the CSE/EE/IEEE

A Survey of Molecular Modeling and Structural Bioinformatics Methods in use at the Nevada Center for Bioinformatics

Computational approaches to the study of Biology and Chemistry have become a mainstay for these disciplines, allowing for the analysis of features and behaviours not accessible from experimental methods. In particular, Molecular Modeling and Structural Bioinformatics methods are increasingly being utilized to search for, design, and analyze data related to the structure-function relationship of Biological macromolecules, such as proteins and DNA. In this talk, I will give an overview of Molecular Modeling methods and techniques and illustrate these methods using examples from several projects at the Center for Bioinformatics.

UNR EE Host

Dr. Sami Fadali

Judith R. Fredrickson

University of Nevada, Reno

Location: SEM 201

Thursday, February 9th, 2006 at 09:30a.m.

Sponsored and organized by the CSE Proposal Colloquium

On the Crossing Number of Complete Graphs

A minimal graph is a graph exhibiting minimum crossings. The crossing number of a graph is the minimum number of edge crossings among the drawings of the graph in the plane. The crossing number problem has applications in the areas of graph layout and circuit design. It is an NP-complete problem. For complete graphs of order n (Kn) the crossing number has been determined only for n < 10.

A conjectured solution for the crossing number of complete graphs of any size was proposed over thirty years ago and to date has not been confirmed or disproven. In the exploration of the illusive crossing number for larger n, we introduce a process to generate minimal Kn from minimal Kn?1. Graph isomorphism is exploited, allowing us to step to the next level with minimal overhead.

UNR CSE Host

Dr. Fred Harris

Dr. Jordan Pollack

Brandeis University

Biography

Jordan Pollack is a Professor of Computer Science and Complex Systems at Brandeis University. His lab is focused on achieving Artificial Intelligence through recreating open-ended co-evolution in a computational process. He has published in many scientific fields, including:

AI
Artificial life
Cognitive science
Educational technology
Evolutionary learning
Neural networks
Robotics
Intellectual property law

He was named in MIT Technology Review “TR10” in 2001 for work on automatically designed robots.

Location: REL 110

Monday, November 14th, 2005 at 4:00p.m.

Sponsored and organized by the CSE/EE/IEEE Colloquium

Recent Progress in Coevolution

For the past decade my students and I have worked on coevolutionary learning, both in theory and in applications such as learning game strategies in Tic-Tac-Toe or Backgammon, solving problems like sorting networks and CA rules, and designing robot bodies and brains.

Coevolution tries to formalize a computational “arms race” which would lead to the emergence of sophisticated design WITHOUT an intelligent designer, or his fingerprints left in the choice of data representations and fitness function. Coevolution often takes the shape of a game tournament where the players who do well replicate (with mutation) faster than the losers. The fitness function, rather than being absolute, is thus relative to the current population. We have had successes, but we find that often, the competitive dynamics lead to winner-take-all equilibria, boom and bust cycles of memory loss, and mediocre stable states where an oligarchy arises which survives by excluding innovation rather than embracing it. Many researchers have proposed algorithmic methods for overcoming these limitations, involving diversity maintenance, memory for elite players, and so forth, but something is wrong if we have yet to have a convincing mathematical or computational demonstration that competition without central government can lead to sustained innovation.

Is there a missing principle, a different mechanism design in which self-interested players can optimize their own utility, yet together the population keeps improving at the game? If so, and if we discover this in the realm of computational games, would it transfer it to human social organization?

UNR CSE Host

Dr. Sushil Louis

Dr. Darrell Whitley

Colorado State University

Biography

Professor Whitley is the Chair of the Department of Computer Science at Colorado State University. From 1993 to 1997, he served as Chair of the Governing Board of the International Society for Genetic Algorithms. In 1999 ISGA merged with the Genetic Programming community to form the International Society for Genetic and Evolutionary Computation. From 1997 to 2002 Professor Whitley served as Editor-in-Chief for the journal Evolutionary Computation published by MIT Press. His areas of interest are artificial intelligence, genetic algorithms, heuristic search, neural networks, and scheduling.

Location: SEM 234

Monday, October 10th, 2005 at 1:00p.m.

Sponsored and organized by the CSE / EE / IEEE Colloquium

CoEvolving Cooperative Teams of Unmanned Aerial Vehicles (UAVS)

We use simulated evolution, specifically Genetic Programming, to evolve behaviors for teams of Unmanned Aerial Vehicles (UAVS). The problem can also be posed of one of evolving adaptive teams of cooperative agents. Agents must display cooperative interactions to carry out collaborative missions in an uncertain and/or hostile environment. Behaviors are controlled by program trees constructed from a set of low-level sensor readings. Evolved program trees are robust to changes in initial mission parameters. The evolved programs are also provably near optimal, compared to optimal time-to-completion under static conditions.

A second question explored in this talk asks if other methods can be used to evolve behaviors similar to those found using traditional Genetic Programming. Experiments probe this question using a variety of benchmark problems and approaches.

UNR CSE Host

Dr. Sushil Louis

Dr. Masoud Nikravesh

University of California, Berkeley

Biography

Dr. Masoud Nikravesh is the Executive Director of the Berkeley Initiative in Soft Computing (BISC) in the Computer Science Division at the University of California, Berkeley. BISC is a world-leading center for basic and applied research in soft computing, computing with words (CW), computational theory of perception (CTP), and precisiated natural language (PNL). Dr. Nikravesh is a leading expert in soft computing, which embraces several disciplines of computing technologies, including:

Neurocomputing
Fuzzy logic
Evolutionary computing (including DNA-based computing)
Probabilistic reasoning

He is a board member of many private and public centers of IT excellence. His achievements have led to front-page news stories in Lawrence Berkeley National Laboratory News, as well as headline news in the Electronics Engineering Times.

Location: SEM 234

Monday, September 19th, 2005 at 2:30p.m.

Sponsored and organized by the CSE / EE / IEEE Colloquium

Intelligent Decision Support and Information Systems: Neuro-Fuzzy-Evolutionary Computing-Based Recognition Technology

Searching database records, ranking the results and making decisions based on multi-criteria queries is central for many database applications and information systems in enterprise used within organizations in finance, business, industrial and other fields. Most of the available systems are modeled using crisp logic and queries, which result in rigid systems with imprecise results. In this presentation, we introduce an intelligent decision support and information systems in enterprise based on evolutionary-based optimization as a flexible tool allowing approximation where the selected objects do not need to match exactly the decision criteria resembling natural human behavior. The model consists of five major modules:

The Fuzzy Search Engine
The Application Templates
The User Interface
The Database
The Evolutionary Computing

The system is designed in a generic form to accommodate more diverse applications and to be delivered as stand-alone software to academia and businesses.

UNR CSE Host

Dr. Gregory Vert

Dr. Masoud Nikravesh

University of California, Berkeley

Biography

Dr. Masoud Nikravesh is the Executive Director of the Berkeley Initiative in Soft Computing (BISC) in the Computer Science Division at the University of California, Berkeley. BISC is a world-leading center for basic and applied research in soft computing, computing with words (CW), computational theory of perception (CTP), and precisiated natural language (PNL). Dr. Nikravesh is a leading expert in soft computing, which embraces several disciplines of computing technologies, including:

Neurocomputing
Fuzzy logic
Evolutionary computing (including DNA-based computing)
Probabilistic reasoning

He is a board member of many private and public centers of IT excellence. His achievements have led to front-page news stories in Lawrence Berkeley National Laboratory News, as well as headline news in the Electronics Engineering Times.

Location: Access Grid Node (SEM 201)

Monday, September 19th, 2005 at Noon

Sponsored and organized by the CSE / EE / IEEE

Human Mind, Intelligent Systems, and Computation: Evolution of Fuzzy Logic and Logic of Vagueness

Inspired by human's remarkable capability to perform a wide variety of physical and mental tasks without any measurements and computations and dissatisfied with classical logic as a tool for modeling human reasoning in an imprecise environment, Lotfi A. Zadeh developed the theory and foundation of fuzzy logic with his 1965 paper “Fuzzy Sets” and extended his work with his 2005 paper “Toward a Generalized Theory of Uncertainty (GTU)-An Outline.” Fuzzy logic has at least two main sources over the past century. The first of these sources was initiated by Peirce in the form he called a logic of vagueness in 1900s, and the second source is Lotfi's A. Zadeh work, fuzzy sets and fuzzy Logic in the 1960s and 1970s.

UNR EE Host

Dr. Sami Fadali

Paul Seidler

Robinson/Seidler, Inc., Henderson, NV

Biography

Paul Seidler has an M.A. in Public Policy from the University of Chicago and a B. A. from the University of Illinois. His interest in nuclear issues arose in the 1970s when he was an undergraduate student and became involved in the antinuclear movement. Mr. Seidler worked on Ted Kennedy's campaign for president and was President of the campus Democratic organization. He later worked for Paul Simon in his successful bid to unseat US Senator Charles Percy.

Mr. Seidler's views on nuclear issues continued to evolve when he worked for the Illinois Department of Nuclear Safety. Unlike Nevada, Illinois has a state-of-the-art, very proactive system for dealing with nuclear waste. Paul Seidler was directly involved in escorting and inspecting spent fuel shipments, but his primary responsibility was setting up a low-level radioactive waste (LLW) disposal facility. He was the first person to successfully identify a "volunteer" community to serve as host for a radioactive waste disposal facility. Several years later he was recruited to work on the nation's first high-level geologic repository.

In the late 1980s, Mr. Seidler moved to Las Vegas to work for Science Applications International Corporation. He was responsible for waste transportation policy issues and intergovernmental relations. He developed an excellent working relationship with the rural communities most impacted by the project and waste transport. These relationships remain strong today. In the mid 90's he formed his own consulting firm with Ace Robison.

Location: SEM 326

Wednesday, May 25th, 2005 at Noon

Sponsored and organized by the IEEE / Nevada Nuclear Energy

Some REAL Yucca Mountain Information

Additional event details will be made available shortly

Dr. Martin Gollery

University of Nevada, Reno Center for Bioinformatics, Department of Biochemistry & Molecular Biology

Biography

Dr. Martin Gollery is the Associate Director of Bioinformatics at the Nevada Center For Bioinformatics. His interests include the representation of Protein Families through the use of Hidden Markov Models. As the former Director of Bioinformatics Research at TimeLogic, he is perhaps best known for his work with accelerating Bioinformatics algorithms such as TeraBLAST and HMMpfam by hundreds or thousands of times through the use of FPGA's. He is an award-winning instructor, popular speaker on the conference circuit, and is frequently quoted by trade journals.

Location: SEM 261

Wednesday, April 13th, 2005 at 1:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Hardware Acceleration in Bioinformatics

Over the past several years, the growth of Biological Databases have consistently exceeded Moore's law, causing the analysis of that data to fall behind. The current popularity of computing clusters has run into several bottlenecks due to the ongoing costs related to power, cooling, floorspace and manpower. The use of FPGAs to provide computing power for the most CPU intensive algorithms can alleviate the need to constantly upgrade the local cluster to enable efficient data mining. This presentation will provide an overview of the current state of the art for Bioinformatics acceleration through the use of FPGAs.

UNR CSE Host

Dr. Gregory Vert

Dr. Brian W. Beck

University of Nevada, Reno Center for Bioinformatics, Department of Biochemistry & Molecular Biology

Biography

Dr. Brian W. Beck is the Associate Director of Molecular Modeling for the Center for Bioinformatics and the Department of Biochemistry & Molecular Biology at the University of Nevada, Reno. He received a BS in Biochemistry from Texas A&M University and a Ph.D. in Biochemistry & Biophysics from Washington State University where he worked with Prof. Toshiko Ichiye on computational analysis (particularly molecular dynamics simulations) of the manner in which electron transfer proteins regulate their redox potentials. His current work involves the use and development of computational techniques for the analysis of the relationship biomacromolecular structure to function, in in particular the way in which solvent affects protein structure.

Location: SEM 261

Wednesday, April 6th, 2005 at 1:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

An Introduction to Methods and Algorithms in Molecular Modeling and Structural Bioinformatics

Computational approaches to the study of Biology and Chemistry have become a mainstay for these disciplines, allowing for the analysis of features and behaviours not accessible from experimental methods. In particular, Molecular Modeling and Structural Bioinformatics methods are increasingly being utilized to search for, design, and analyze data related to the structure-function relationship of Biological macromolecules, such as proteins and DNA. Instructor this talk, I will define Molecular Modeling, discuss the more common models and their relationship with experimental data, and then discuss common methods such as Drug Design, Comparative Protein Modeling, and Structural Genomics. I will then briefly mention several projects at the Center for Bioinformatics and the hardware & software we utilize to undertake them.

UNR CSE Host

Dr. Gregory Vert

Mr. Jeff Elpern

Reno/Tahoe Software Quality Institute

Biography

Jeff Elpern is a high-tech executive and entrepreneur. Currently he is the V.P. of Software for a Silicon Valley telecommunications component manufacturer, founder and CEO of the Software Quality Institute (SQI, Inc.) based in Reno, and founder of the non-profit Open Source Nevada. He was founder of two startups and on the executive teams for two other start-ups. Earlier in his career, he ran the largest quantitative marketing operation on Madison Avenue and was part of the Lee Iacocca turn-around team at Chrysler. He is a native Nevadan who completed his undergraduate work at UNR and received a Masters of Science in Quantitative Analysis from Carnegie Mellon University.

Location: SEM 234

Monday, March 14th, 2005 at 2:30p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

A Unified Theory of Software Quality: Economic Quality, not Code Quality

The presentation introduces the Unified Theory of Software Quality (UTSQ) - a conceptual framework and econometric model focused on Economic Return as the only true measure of software quality. The inability of current quality models - SEI CMM or ISO for example - to explain the successful development strategies of most real-world software companies is discussed and the need for a more encompassing theory is presented. It is argued that a User's Economic-Return quality metric - as opposed to code quality metrics such as defect density or Six Sigma - provides a basis for a theory that can predict rational behavior across all development efforts. The engineering behavior of game developers, productivity application developers, telco embedded controls developers, weapons developers, and nuclear power plant controls developers seems radically different. However, this presentation of UTSQ shows all are acting from the need to maximize User Economic Return. UTSQ is successful in predicting development strategy because the econometric framework directly ties a firm's revenue to the software's Economic Quality.

UTSQ puts the user at the center of all quality metrics. The presentation introduces the "user in the loop" concept of quality. The presentation identifies a user as a critical component of Economic Quality. The presentation will also cover these aspects of UTSQ: task fixed value, difference between value and return, impact of catastrophic failure potential, brand vs. product, and user's response/cost of task stimuli. Actionable heuristics - derived from the UTSQ framework - are presented. For example, the economic quality is always improved by moving a sub-task from ResourceWare to KnowledgeWare or from PeopleWare to Software (the presentation will define these terms). Or, the development team that successfully pioneered a new application category will most likely be ineffective as the category matures because the focus moves from adding economic value to reducing economic costs.

UNR CSE Host

Dr. Sergiu Dascalu

Dr. Mark Wessinger

University of Nevada, Reno Department of Psychology

Biography

Professor Wessinger received his Ph.D. in Neuroscience from the University of California, Davis in 1995. He mapped the residual visual capabilities of patients with partial blindness. This was the beginning step in his quest to understand how our brains process the enormously complex environment. Following his Ph.D., Wessinger remained at UCD to complete a post-doc in functional magnetic resonance imaging (fMRI). This technique allows one to "see the brain in action." The projectWessinger headed mapped the regions of the brain involved in auditory perception. In 1996, Wessinger accepted a position as a Research Associate in the GeorgetownInstitute of Computational and Cognitive Science where he extended this research by using fMRI to map regions of the brain involved in speech perception. His next move, in 1998, was to Dartmouth as a Visiting Research Professor. There he began teaching, as well as continuing his research using fMRI to explore auditory processing by becoming involved in a study investigating musical perception. He also pursued investigations of residual vision by working with additional patients, as well as modifying the paradigm to investigate issues of awareness in healthy individuals. In 2001, Wessinger moved to Gettysburg College where he taught several psychology courses, implemented a cognitive neuroscience research program involving undergraduate students in fMRI, and helped develop a neuroscience minor. His research focus was investigating multi-modal processing of objects and animals. This involved collecting visual and tactile data in the laboratory and the fMRI machine. Seeking to continue quality teaching at the undergraduate level, as well as re-establish a stronger and more diverse research program that involves both graduate and undergraduate students, Wessinger joined the Experimental Group in the Psychology Department at UNR during the fall of 2004. He is currently implementing a cognitive neuroscience program, involving graduate and undergraduate students. This fall he taught, Introductory Psychology and Cognitive Psychology. This spring, he is teaching Experimental Psychology, a core course for majors that focuses on research methods used in the behavioral sciences.

Location: MS 321

Thursday, March 3rd, 2005 at 4:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Multimodal Investigations of Perceptual and Conceptual Processing

In this talk I will describe how I have used functional magnetic resonance imaging (fMRI) to investigate various aspects of perceptual and conceptual processing in multiple modalities. After giving a brief explanation of the basis of magnetic resonance imaging (MRI) and fMRI, I will describe how I came to use the technique in my research. My first study was using fMRI to demonstrate tonotopic organization in human auditory cortex. Then I will segue into how this lead to my involvement in a project using fMRI to explore aspects of human speech perception, which lead to investigations of differential processing of musical sequences in musicians and non-musicians. These studies make a nice perceptual-to-conceptual progression in the auditory domain.

Subsequently, I will present a similar progression in the visual and tactile domains, beginning with rudimentary demonstrations of retinotopic & somatopic organization and ending with comparisons of conceptual processing of animals and objects in the visual and tactile domains. I will conclude by relating these seemingly disparate studies with the overall goal of understanding how we act, react, and ultimately (hopefully :)) survive within our incredibly information-rich environment.

UNR CSE Host

Dr. George Bebis

Dr. Jeffrey B. Mulligan

NASA Ames Research Center

Biography

Dr. Jeffrey B. Mulligan received the A.B. degree in physics from Harvard College in 1980 and the Ph.D. degree in psychology from the University of California at San Diego in 1986. He is currently a computer engineer at the NASA Ames Research Center, where his primary research interests are human spatial vision (with emphasis on motion perception and eye movements) and image processing. He is a member of the Optical Society of America, the Association for Research on Vision and Ophthalmology, and the Society for Information Display.

Location: MSS 216

Wednesday, February 23rd, 2005 at Noon

Sponsored and organized by the Pyschology / CSE Seminar

The Collection of Gaze Data in Operational Environments

Gaze tracking measures can provide information about the spatial locus of attention of a behaving agent. This talk reviews three technologies I have been developing to allow the collection of gaze data in operational environments. The first involved head-based gaze tracking of a working air traffic controller, using low-quality video data from an overhead camera. Using a combination of automatic and operator-assisted procedures, we constructed a 3-D graphical model of the subject's head, which was subsequently used to match the pose in each frame of an extended sequence, allowing a crude categorization of gaze into regions such as the tower windows and computer displays. I will also present prelimnary results from a 3 camera system designed to accurately track the gaze of a freely-moving workstation user. An active illumination system coupled with a stereo imaging system is used to track the positions of the eyes in 3-D space; the results are then used to control a third, steerable narrow-field camera used for gaze estimation. Finally, I will describe a study examining the looking behavior of helicopter pilots attempting to fly precision routes with the aid of a Global Positioning System (GPS) receiver, illustrating some of the unique challenges which arise in outdoor data collection.

UNR CSE Host

Dr. George Bebis

Dr. Krzysztof Podgorski

University of Nevada, Reno Department of Mathematics and Statistics

Biography

Dr. Krzysztof Podgorski is an Associate Professor with the Department of Mathematics and Statistics, UNR. In 1991 he received a Ph.D. in Mathematics from Wroclaw Technical University, Poland and in 1993 a Ph.D. in Statistics from Michigan State University. Between 1994 and 2004 he was a faculty at Indiana University - Purdue University, Indianapolis. Dr. Podgorski's main research interests are in Stochastic Modeling and Mathematical Statistics.

Location: SEM 234

Tuesday, February 22nd, 2005 at 4:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Monte Carlo Inverse

The inverse problem is defined as finding a solution to a system of functional equations that do not have a unique solution. This problem is very common in physics and engineering sciences - for example, the problem of mapping the sea-bottom from various measurements taken from a ship. Image recovery can also be formulated as such a problem. In this seminar, one of the possible non-linear solutions to this problem, called the Monte Carlo Inverse, will be presented. Criteria for the existence of solutions and for convergence of Monte Carlo algorithms to such solutions will also be discussed. The method presented has applications in mathematical finance and machine learning.

UNR CSE Host

Dr. George Bebis

Mr. Jeff Helfand

Nevada Interactive

Biography

Mr Jeff Helfand is a Business School graduate from San Jose State University. Having spent over three decades growing up and working at several Silicon Valley start-ups, Jeff relocated to Reno, NV in 2001. Mr. Helfand has started up several electronic media companies, was the Project Manager of web development for mega online retailer Fogdog Sports, and served as Vice President of eCommerce, Sales and Marketing for ConsumerReview.com.

Location: SEM 234

Wednesday, February 16th, 2005 at 2:30p.m.

Sponsored and organized by the CSE Seminar

Squirrel Squabble: A PC Game for the "Casual Gamers" Market

Recently, in an effort to help kick start a lack of non gaming technology ventures in Reno, Mr. Jeff Helfand formed a partnership with his NevadaInteractive.com consultancy business and the local educational community to form an internship program for interactive entertainment software development. The result was a first time ever entry of a PC video game developed primary by UNR students into the global Independent Games Festival student showcase competition.

The game, Squirrel Squabble, was selected as the winner of the 2005 IGF Student Showcase amongst ninety seven other entries from universities and colleges around the world. In addition, Squirrel Squabble will be debuted to the professional development community at the 2005 Game Developers Conference in San Francisco, CA. At this event, the students who created the game will have an opportunity to pursue commercial distribution of the game with showcase sponsors/publishers such as Microsoft, AOL Sony and Atari.

UNR CSE Host

Dr. Sushil Louis

Dr. Alice O'Toole

The University of Texas at Dallas (UTD)

Biography

A part of the IGT Distinguished Speaker Series

Location: WRB 2025

Tuesday, January 18th, 2005 at 4:00p.m.

Sponsored and organized by the Psychology / CSE Seminar

View-Independent Face Recognition Using Adaptation

People identify a human face more accurately following adaptation to a synthetically created "anti-face" with "opposite" features. (Leopold et al., 2001). This opponent "identity adaptation" originates within perceptually salient feature dimensions of faces. Selective adaptation to the global elements of face shape (Webster & MacClin, 1999) and the natural categorical dimensions of gender, ethnicity, and facial expression (Webster et al., 2004) facilitate the perception of faces with opposite values on these feature dimensions. In the present study, we extend the use of adaptation to investigate the nature of the underlying visual representation of facial identity at levels of neural processing that support view-independent recognition. Previous experiments have demonstrated that face adaptation survives two-dimensional scaling and shifts in retinal position, placing the locus of the effect in high-level visual areas, beyond those with strict retinotopic organization. We show here that opponent identity adaptation also survives a change in 3D viewpoint, indicating that it can tap view-independent face encoding mechanisms. To examine the nature of the visual information underlying view-independent face adaptation, we used opponent-based facial identity adaptation, in combination with 3D morphing of laser scans. We show that the three-dimensional shape and the surface reflectance information in faces can be adapted separately. Surprisingly, both types of adaptation transfer across viewpoint. These findings indicate that neural representation of faces includes both shape and reflectance information in a form that generalizes across changes in three-dimensional viewpoint. This suggests a novel theory of the visual information retained in high level face representations. The presentation covers the work of Dr. O'Toole (UTD), Fang Jiang (UTD), and Volker Blanz (Max-Planck-Institut für Informatik).

UNR CSE Host

Dr. George Bebis

Dr. Allen Brady

Professor Emeritus of Computer Science
University of Nevada, Reno

Biography

Allen Brady began his professional life as a physicist with the Ionoshperic Physics Division of the Central Radio Propagation Laboratory of the (then) National Bureau of Standards in Boulder, Colorado. His later experience there applying computers to data collection and analysis led him to refocus his graduate study in mathematics to the budding area of computer science. After a brief start in teaching computer science following graduate school, he became involved in computing center administration for thirteen years before returning to the full time teaching of mathematics and computer science.

In 1990 he was a member of the founding faculty of the Department of Computer Science (now the Department of Computer Science & Engineering) here at the University of Nevada, Reno. He served as chair of the department from 1991 until his retirement in 1994. In addition to the University of Nevada, Reno, he has held teaching positions at the University of Notre Dame and at the American University in Cairo. He has a B.S. degree in Engineering Physics from the University of Colorado, a M.S. in Mathematics from the University of Wyoming, and a Ph.D. in Mathematics from Oregon State University.

Brady's interest in the Busy Beaver Game began with his discovery in 1964 of the "champion" four-state Turing machine using a Scientific Data Systems SDS-920 computer that was located at a research laboratory in Beaverton, Oregon. He finally completed the solution to the four-state problem in 1974 by programming a computer to search heuristically among some 6,000 remaining four-state Turing machines for characteristics of the individual machines which would then lead the computer to carry out proofs by mathematical induction that they would indeed never halt when started on a blank input tape. He also contributed a chapter on the Busy Beaver Game to Rolf Herken's 1988 book celebrating "The Universal Turing Machine A Half Century Survey' (Oxford University Press, 1988). A computer program written in 1986 to generate data for that chapter was used in his recent effort to discover new lower bounds for the three-state by three-symbol case that he announced on the Internet on November 15, 2004.

He recalls giving two previous colloquium lectures on the campus. In 1975, after completing the construction of his own microcomputer from a kit (Altair 8800), having spent more than 70 hours soldering individual parts and wires to printed circuit boards, he gave what was the first lecture on microcomputers at the . It was attended by an overflow audience and had to be moved to a larger room. In 1984 he gave a lecture review of Andrew Hodges' newly published biography "Alan Turing: The Enigma' (Burnett Books,1983) to a very small but intensely interested audience.

Location: REL 109

Tuesday, November 30th, 2004 at 3:30p.m.

Sponsored and organized by the CSE Seminar

Tibor Rado's Busy Beaver Game - An Old Computer Science Problem for a New Kind of Science

The study of computability began as an obscure area in mathematical logic more than ten years before the advent of the modern computer. It concerns computational universality (what mechanism can do) and computational unsolvability (what mechanism or machines cannot do). The limitations on what cannot be done by any sort of "computing machinery" have immediate consequences that weigh heavily on the success, failure, and reliability of the increasingly complex systems being designed by computer scientists and engineers. These limitations likewise impose great difficulties for our complete understanding of the mechanisms at work in biology, including ecological systems and evolution. In recent years there has also been speculation about simple logical mechanisms of a "computationally universal" nature underlying the behavior of physical matter in the universe.

For students, computability is typically relegated to an advanced "theory" course in Computer Science where its relevance to their real concerns seems to be lost. The subject is in fact completely accessible at a very elementary level where the power of universality can be immediately appreciated and the implications of unsolvability can be made clear. The Busy Beaver Game was invented by Tibor Rado as an instructional device to introduce computability via Turing machines to his mathematics students. The goal of the game was to find the largest number of marks that a Turing machine with a fixed number of states and a two-symbol alphabet (a mark and blank) could leave on an all blank input tape before coming to a halt. The number of marks left by a machine would be its "score", and the machine (or the student discovering it) would be the designated "champion" of the contest. This simple but challenging game turns out to describe a noncomputable function Sigma(k) defined as the maximum score that can possibly be achieved by some Turing machine of k-states. Related to Sigma(k) is another function S(k) defined as the maximum number of steps or "shifts" that can be taken by such a k-state Turing machine before halting. S(k)>=Sigma(k), and Rado showed that these functions must grow more rapidly than any computable function!

The Busy Beaver problem was long ago solved for k=2, 3, and 4 for Turing machines operating with k states and two symbols (the answers being maximum scores of 4, 6 and 13 respectively and correspondingly a maximum of 6, 21, and 107 steps possible before halting). When the number of states is raised from 4 to 5, the problem seems to explode. It is now known (since 1989) that the best possible "score" for a five-state machine must be at least 4,098 achieved by a Turing machine which carries out 47,176,870 steps to achieve that result.

If one steps back to the case for k=3 -- the 3-state Turing machines with two symbols achieving a maximum score of 6 and a maximum number of steps before halting of 21 -- and examines the effect of adding another symbol instead of another state, a complete surprise is encountered. While the problem remained manageable after adding only one state (Sigma(4) = Sigma(4,2) = 13 and S(4) = S(4,2) = 107), a new study shows that Sigma(3,3) >= 5,600 and S(3,3) >= 29,403,894.

The surprising results of this recently completed study will be presented. It involved utilization of programs running concurrently on as many as 17 networked computers each running for as long as 12 hours at night and on weekends for a period of nearly two weeks. More than 28 million 3x3 Turing machines were generated and each machine was allowed to run for up to 50 million steps while confined to work within a tape of 20,000 squares in length. (What began as a quick look for easy answers a month earlier using an old program on a home computer rapidly turned into an exhausting effort.)

The problem is ideally suited for further study on a large parallel cluster where the extreme and unpredictable nature of the load balance using parallel execution might be more easily accommodated. How such an approach might be made will be discussed briefly.

Finally, the Busy Beaver problem is related to the class of problems, including cellular automata, presented by Stephen Wolfram in his book A New Kind of Science (Wolfram Media, 2002). In fact the Busy Beaver Game is discussed in the book as a related problem. Wolfram proposes that this area of study should be treated as a separate discipline ("NKS") that becomes part of the education of all scientists. Wolfram's position will be examined. Likewise, from the computer science perspective the design and use of critical computer systems (including even voting machines) with little appreciation by the designers for unseen consequences will be commented on from the computer science and engineering perspective. Moving slightly away from computers, the question will be raised about what appears to be the "black box" automata approach in "genetic engineering": fiddle with the input a little bit and see what comes out -- never mind that we have no idea what's inside the box or that worse yet we might inject the output into other more complex systems about which we know even less! Finally, a plea will be made for an early introduction of computability issues into the computer science curriculum to encourage a little more humility in the approach to the design of software.

UNR CSE Host

Dr. Fred Harris

Dr. Brian Gerkey

Stanford University

Biography

Dr. Brian Gerkey is a postdoctoral research fellow in the Artificial Intelligence Lab at Stanford University. Dr. Gerkey received his Ph.D. in Computer Science from the University of Southern California (USC) in 2003, his M.S. in Computer Science from USC in 2000, and his B.S.E. in Computer Engineering, with a secondary major in Mathematics and a minor in Robotics & Automation, from Tulane University in 1998. His primary research interest is the development, evaluation, and analysis of principled (yet practical) coordination algorithms for multi-robot systems. He is also a founding and co-lead developer on the Player/Stage/Gazebo project.

Location: REL 110

Thursday, October 28th, 2004 at 09:30a.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

The Player/Stage/Gazebo Project: Open Source Tools for Robotics Research

Research in robotics requires a significant amount of hardware and software infrastructure. While the rise of commercial robot vendors (e.g., ActivMedia, iRobot, Segway) has standardized the hardware in use in many labs, there is no equivalent standard for the software. As a result, the proverbial wheel has been reinvented many times by many robotics researchers, who have unnecessarily re-implemented everything from device drivers for common hardware to simulations of common robots.

I will present a solution to this problem: the Player/Stage/Gazebo (P/S/G) project. This project produces powerful, flexible, Open Source tools for developing and experimenting with mobile robots and sensors. Player is a network server that provides a clean, abstract interface to a wide variety of robot hardware and software; Stage and Gazebo are simulators that provide populations of virtual Player devices. P/S/G has a world-wide community of users, from undergraduates doing class projects in simulation, through graduate students using teams of robots, to multiple-institution DARPA-funded research teams using Player for systems integration.

I will discuss the key ideas that characterize the P/S/G project: the character device model, the interface/driver model, the client/server model, and the Open Source model. By making effective use of these well-understood ideas, we have created a successful and popular software suite (over 16,000 downloads to date) that has been described as "one of the leading candidates for an open, standardized robot control and simulation interface." I will also discuss future directions and enhancements that are in the works. For more information (and, of course, the software itself), please visit: The Player/Stage Project.

UNR CSE Host

Dr. Monica Nicolescu

Dr. Kenneth W. Tobin

Oak Ridge National Laboratory

Biography

Dr. Kenneth W. Tobin is a Corporate Research Fellow and Group Leader of the Image Science and Machine Vision Group at the Oak Ridge National Laboratory, Oak Ridge, Tennessee. The group performs applied computer vision research and development in industrial inspection and metrology, biomedical imaging, and national security. He performs research in non-destructive test and analysis, image processing, and image-based metrology for automation and process characterization. He has authored and co-authored over 120 publications and he currently holds seven U.S. Patents with two additional patents pending in the areas of computer vision, photonics, radiography, and microscopy. Dr. Tobin is a Fellow of the International Society for Optical Engineering (SPIE) and a Senior Member of the Institute of Electrical and Electronics Engineers (IEEE). He has a Ph.D. in Nuclear Engineering from the University of Virginia, Charlottesville, Virginia, and an MS in Nuclear Engineering and a BS in Physics from Virginia Tech., Blacksburg, Virginia.

Location: REL 110

Monday, October 18th, 2004 at 4:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Applied Computer Vision Research at the Oak Ridge National Laboratory

The Oak Ridge National Laboratory (ORNL), Oak Ridge, Tennessee, is the U.S. Department of Energy's largest science and energy laboratory. Established in 1943 as part of the Manhattan Project, today ORNL supports the nation with a peacetime mission of scientific competencies including neutron science, energy research, high performance computing, complex biological systems, advanced materials, and national security. The Image Science and Machine Vision Group was established in 1987 to conduct applied research to develop technologies that provide human-like decision making capabilities for computers and robots. Today the group performs research in a variety of areas such as image-based inspection and metrology, scene analysis and comprehension, and archival image indexing and management. In this presentation Dr. Tobin will review several areas of their research including: small animal x-ray computed tomography and SPECT imaging, semiconductor wafer and lithographic mask inspection and metrology by interferometric imaging, and applications of content-based image retrieval for managing large image repositories of industrial, biological, and geographic data.

UNR CSE Host

Dr. George Bebis

Dr. Darko Koracin

Desert Research Institute, Reno, NV

Biography

Dr. Darko Koracin is a research professor in the Division of Atmospheric Sciences of the Desert Research Institute. He has received B.Sc. and M.S. degrees in Geophysics and Meteorology from the University of Zagreb, Croatia, and a Ph.D. degree in Atmospheric Physics from the University of Nevada, Reno. He investigates the properties and evolution of atmospheric flows over both complex terrain and the ocean through the development and application of high-resolution mesoscale meteorological models. His expertise includes modeling and observational studies of the transport and dispersion of atmospheric pollutants and tracers using and developing dispersion models of various complexities. Dr. Koracin is an instructor at the UNR Atmospheric Sciences department and he is mentoring students in M.S. and Ph.D. programs.

Location: REL 110

Thursday, September 30th, 2004 at 3:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Simulations of Atmospheric and Dispersion Processes in Complex Terrain and Over the Ocean

The behavior of the atmospheric and dispersion processes can be described as a mathematical formalism by a set of nonlinear differential equations whose solution is yet to be found. Consequently, we design atmospheric and dispersion computer models in which numerical techniques are used to obtain the solution for the set of equations. We have been using atmospheric and dispersion models for a variety of basic and applied research studies focused on real-time weather forecasting, air quality, coastal dynamics, weather modification, wind energy, and tracer experiments. Due to the complexity of the models, significant computer efforts in execution and post-processing are required. This requirement was recently emphasized by the need to execute simultaneously coupled atmospheric and oceanic as well as hydrological models. Some of the issues related to code parallelization, visualization, and computer requirements will be discussed during the presentation.

UNR CSE Host

Dr. George Bebis

Dr. Graham Kendall

Senior Lecturer
School of Computer Science and Information Technology
University of Nottingham

Biography

Dr. Graham Kendall is a senior lecturer in the School of Computer Science and Information Technology at the University of Nottingham. He is a member of the Automated Scheduling, Optimization and Planning (ASAP) research group. Before undertaking his academic career, he spent over 15 years in the IT industry where he managed upwards of 50 people and managed a variety of small, medium and large scale IT projects. He is a member of the EPSRC (Engineering and Physical Sciences Research Council) peer college and an associate editor of three international journals. Dr Kendall is the principal investigator on four active awards from and co-investigator on eight other external awards. Dr. Kendall has attracted over £1.6M in external funding in a little over four years. He has published over 40 refereed papers in international journals and conferences, is the editor of six books and has served on over 25 program committees since 2000.

Location: SEM 203 - Access Grid Node

Thursday, June 24th, 2004 at 10:00a.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Hyper-heuristics: An Emerging Search Technology

We introduce an emerging methodology in search and optimization. The aim of this approach, which has been termed hyper-heuristics, is to raise the level of generality at which optimization systems can operate. The hope is that hyper-heuristics will lead to more general systems that are able to handle a wider range of problem domains than current (meta-) heuristic approaches, which tend to be customized to a particular class of problems or even specific problem instances. Hyper-heuristics are broadly concerned with intelligently choosing the right heuristic (or algorithm) at each decision point to produce solutions which are good enough, soon enough, and cheap enough. In this talk we will give a brief history of this emerging area and discuss the progress made to date.

Jae-Khun Chang

Associate Professor
Dept. of Computer Science
Hanshin University

Location: MS 321

Friday, April 30th, 2004 at 3:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Automatic Intellectual Tracking/Recognition of Multi-targets

The recognition and tracking technology for multiple moving objects in the road is presented, especially for cases where objects cross each other. The technologies developed in this research include:

Detection of moving objects and distinguishing them from a stationary background
Tracking multiple objects using 2D and 3D information processing
Image stabilization and data fusion

In object recognition, background generation process classifies pixels into background and object pixels. For image stabilization, a line-based approach where lines are formed connecting detected corners and matched across frames is developed. For data fusion, image fusion algorithms combining data from multiple sensors are developed. The technology using a traffic characteristics measurement complex is tested during the final experimental stage of the research. The developed technologies allow traffic monitoring without any road modification and will work with all type of roads.

Dr. Fuhui Long

Duke University

Biography

Fuhui Long is currently a research associate at the Center for Cognitive Neuroscience, Duke University. Her research interests include image processing, computer vision, machine learning and human visual perception.

Location: LME 316

Wednesday, April 21st, 2004 at 4:00p.m.

Sponsored and organized by the CSE Seminar

Scene Structure Categorization

Psychophysical studies have shown that scene structures play important roles in human vision. For instance, human subjects are able to identify scenes without detailed local information of objects, as long as the global relations among large-scale structures in the scenes are preserved. In addition, human percepts of visual targets are greatly influenced by the spatial organization and complexity of their contexts. Motivated by these evidences in human vision, a scene structure categorization approach is presented here. The gist of the idea is to characterize scene structure by the spatial organizations of the regions of different spatial frequency characteristics, which provides cues on how objects in different distances are organized in the scene.

The approach contains four steps:

A spatial frequency map is computed to represent the spatial frequency variations in the scene
The spatial frequency map is segmented into several relatively uniform regions using GMM and EM methods
Represent the spatial relationships among regions and generate similarity matrices for a group of scenes
Spectral clustering is applied to categorize the scene images

The preliminary results show that by setting up the association between regions of different spatial frequency and their contexts, this approach appears to produce encouraging results in categorizing scenes into semi-semantic groups.

David Loeb

President
Excelerate Software

Biography

David Loeb is the Founder and President of Excelerate Software. In the 10 years prior to founding Excelerate, Mr. Loeb was Vice President of Professional Services and Chief Technical Officer for IntelliCorp, Inc. in Silicon Valley. As Vice President of Professional Services at IntelliCorp, he managed the delivery of solutions spanning eCommerce, Customer Relationship Management, and Configure-to-Order. Mr. Loeb has held management positions in consulting and development at AICorp and Enterprise Software. He began his career at IBM's Research Lab. Mr. Loeb studied computer science at the University of Illinois, and is the author of two software patents.

Location: SEM 234

Wednesday, April 21st, 2004 at 2:30p.m.

Sponsored and organized by the CSE Seminar

Personalizing Web Content

Mr. Loeb will discuss the technology utilized to personalize an application or website to each individual's needs. The application of personalization technology for electronic commerce and electronic marketing will be discussed.

Dr. Gregory Vert

Assistant Professor
University of Nevada, Reno

Biography

Dr. Vert received his undergraduate degree, with specialization in geographic information systems from the University of Washington, a Master's degree from Seattle Pacific University in information systems management and a Ph.D in computer science from the University of Idaho. He has worked in industry for 14 years for companies such as IBM, American Express and Boeing, and has taught computer science at the University of Idaho, Portland State University, and the University of Phoenix. His research interests include database and spatial data management, data visualization, computer security and bioinformatics.

Location: Getchell Library Projection Room

Tuesday, April 20th, 2004 at 2:00p.m.

Sponsored and organized by the CSE / IEEE Seminar

A Fuzzy Based Method for Classifying Semantically Equivalent Spatial Data Sets in Spatial Database Queries

Geographic information systems are becoming more important in decision and analysis processes. It is useful to retrieve all GIS data sets that are related to a given geographic theme. At present there are no known methods for implicitly linking data sets so that all related sets are returned for a given spatial query. A new method is being developed that employs fuzzy logic and a geometric glyph that can be applied to the automatic linking and retrieval of spatially related data sets in a query.

Dr. Sanjay Kumar Madria

Department of Computer Science
University of Missouri-Rolla

Biography

Sanjay Kumar Madria received his Ph.D. in Computer Science from Indian Institute of Technology, Delhi, India in 1995. He is an Assistant Professor, Department of Computer Science, at University of Missouri-Rolla, USA. Earlier, he was a visiting Professor at Purdue University, West Lafayette. He has published papers in leading Journals and conferences in the areas of web warehousing, mobile databases, nested transaction management and performance issues. He is a co-author of a book "Web Data Management: A Warehouse Approach" published by Springer in 2003. He guest-edited WWW Journal and Data and Knowledge Engineering Sp. Issues on Web data management and Data warehousing. He was an invited keynote speaker in Annual Computing Congress in Oct.99 in Canada and an invited speaker in Conference on Information Technology, 2001. He has chaired conferences and workshops in the area of web and internet computing. His research has been supported by NSF, DOE, and UM research board grants. He is an IEEE senior member.

Location: REL 109

Tuesday, March 30th, 2004 at 3:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

DiffXML - Detecting Changes in XML Data

With the role of XML getting more and more important in Web applications, storing XML data in different data models is much more interesting. Most of the solutions use relational model for storing XML data. Users are often not only interested in the current documents on Web but also in changes to learn about the evolution of the Web. The change detection methods presented so far uses tree comparison approaches. In this talk, I introduce a method to map XML files to relational data model. The method parses XML files as DOM trees and stores value and path information for each node in a single relational table. Then I will present an algorithm called "DiffXML" which uses SQL operations to detect changes such as move, insert, delete and updates between two versions of an XML file stored in a relational database. The value and path information for XML files are used to detect differences. I will present the performance results of DiffXML compared with some existing prototypes including a commercial XML change detection tool.

Dr. Konstantinos Veropoulos

Biography

Konstantinos Veropoulos was born in Athens, Greece. He has completed a BSc in Computer Science & Engineering at the American University of Athens in Greece, a MSc in Parallel Computer Systems at the University of the West of England and a Ph.D. in Machine Learning and Machine Vision at the University of Bristol in United Kingdom. He is a young researcher in the theory and applications of Artificial Neural Networks and Support Vector Machines and has published and presented his work in a number of key international conferences and journals in this area. His research interests include:

Interactive computer graphics
Image processing, analysis and recognition
Signal processing
Medical diagnostics
Bioinformatics
Machine learning using Artificial Neural Networks and other learning methods such as Support Vector Machines, Decision Trees, Query Learning and Fuzzy Logic.

Location: LME 316

Wednesday, March 24th, 2004 at 4:00p.m.

Sponsored and organized by the CSE Seminar

Machine Learning and Medical Imaging: A Case Study on the Diagnosis of Tuberculosis

The Automated Diagnosis of Tuberculosis (TB) from photomicrographs of sputum smears is an example of a machine learning and vision application that has shown a very promising performance in a preliminary investigation that took place in the University of Bristol, UK. Image processing and analysis techniques have been used to highlight and extract features of tuberculosis bacteria in photomicrographs. Support Vector Machine (SVM) methods were applied for the identification of these bacteria and the results compared to more conventional machine learning methods such as artificial neural networks with backpropagation and scaled conjugate gradient learning algorithms.

The results and conclusions of this investigation not only show that automated diagnosis of TB is feasible, but also that SVMs have the properties to be successfully applied in similar medical problems and provide a beneficial service to the medical community.

Dr. Walter Dosch

Director
Institute of Software Technology and Programming Languages
University of Luebeck, Germany

Biography

Prof. Walter Dosch received a Ph.D. in Computer Science from the Technical University of Munich. Since 1988 he has been Professor of Computer Science and Artificial Intelligence at the University of Augsburg. Since 1996 he is Full Professor and Director of the Institute of Software Technology and Programming Languages at the University of Luebeck. During the term of office 2000-02, he served as Dean of the Faculty of Technology and Sciences.

His research interests comprise software and system development, foundations of software technology, in particular formal methods, specification and transformation. He contributed over 100 publications to conference proceedings and journals. Prof. Dosch served as the program resp. conference chair of several international software engineering conferences and coauthored two introductory textbooks to computer sciences.

Location: REL 109

Tuesday, March 23rd, 2004 at 09:00a.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Functional Specification and Refinement of Interactive Systems

Interactive systems are composed of software and hardware components of different kind. The components communicate asynchronously by exchanging information along connecting channels. Elementary components can be combined into composite components following different composition styles.The specification, systematic design and correct implementation of interactive systems need sound foundations for describing the interface, the behavior and the structure of components.

The behavior of a (non)deterministic component can be characterized by a function (relation) between the communication histories for input and output. A communication history, for short a stream, records the finite or infinite succession of messages of specified type passing through the component's interface. The stream-based description results in a functional model of interactive systems supporting compositional design methods.

The lecture series presents a tutorial on the functional specification, design and implementation of asynchronously communicating components in distributed systems. On the specification level, we concentrate on functional requirements defining the input/output behavior. During validation we are interested in the component's safety and lifeness properties. Among the design issues, we present compositional techniques for the stepwise refinement of the component's behavior, interface and structure. On the implementation level, we discuss the introduction of data states and control states.

The approach is explicated presenting various processing, memory, control and synchronization components. The accompanying examples illustrate different description styles using recursive functions, logical assertions and graphical representations. We also relate the history-based approach to pragmatic engineering techniques based on state transition tables and abstract state machines.

The lecture series addresses computer scientists interested in the foundations of software and systems engineering.

Dr. Walter Dosch

Director
Institute of Software Technology and Programming Languages
University of Luebeck, Germany

Biography

Prof. Walter Dosch received a Ph.D. in Computer Science from the Technical University of Munich. Since 1988 he has been Professor of Computer Science and Artificial Intelligence at the University of Augsburg. Since 1996 he is Full Professor and Director of the Institute of Software Technology and Programming Languages at the University of Luebeck. During the term of office 2000-02, he served as Dean of the Faculty of Technology and Sciences.

His research interests comprise software and system development, foundations of software technology, in particular formal methods, specification and transformation. He contributed over 100 publications to conference proceedings and journals. Prof. Dosch served as the program resp. conference chair of several international software engineering conferences and coauthored two introductory textbooks to computer sciences.

Location: SEM 234

Monday, March 22nd, 2004 at 2:30p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Bridging System Views in Software Design

Software development has matured from heuristic practice to an engineering discipline. Nowadays software engineers can benefit from a solid stock of basic research addressing specification, modelling, design and implementation techniques for sequential, concurrent, distributed and real time systems. A general system model should provide different system views, in particular a data model, a state transition model, an architectural model, and a process model. Software development must safely bridge system views on different levels of abstraction when implementing a specified behaviour.

Against this general background, we present a functional model for communicating components in distributed systems along with refinement transformations relating external and internal views. In the external view, a component forms a black box interacting with the environment by message passing. The internal view discloses the component's state when implementing the specified behaviour by state transition model.

We discuss formal methods for systematically transforming the input/output-oriented external view into a state-based internal view and for further refining these system views. We conclude with a critical assessment of applying formal methods from academics to industrial software engineering.

Dr. Fengliang Xu

Computer and Information Science Department
Ohio State University

Biography

Dr. Fengliang Xu received his Bachelors of Science Degree in Electrical Engineering in 1996 from the University of Science and Technology of China, and his PhD from the Ohio State University (OSU). He has been involved with the OSU Mapping and GIS Lab performing work in Mars Mapping using Robot Vision. Also, he is a collaborator to the Computer and Information Science Dept of OSU and Fundamental Research Dept. of Honda R&D America in human recognition.

Location: SEM 234

Tuesday, March 16th, 2004 at 4:00p.m.

Sponsored and organized by the CSE Seminar

Mars Mapping & Localization with Robot Vision, Night Vision, and Depth Vision

Mapping and localization techniques in the current Mars Exploration Rover (MER) mission is to generate digital elevation model (DEM) and orthophoto using stereo images from the Navcam and Pancam as well as descent images and orbital images, and matching of these orthophotos to provide precise locations of the rovers. The pipeline is: automatic feature (interest point) extraction and matching, triangulation, mapping, localization, landmark (rock and hollows) extraction and modeling. In night vision, pedestrians are detected using support vector machines (SVM) as well as motion estimation from stereo images. In Depth vision, human and objects including luggage are detected and modeled from depth and grayscale images.

Ibrahim Gokcen

Department of Electrical Engineering and Computer Science
Tulane University

Biography

Mr. Ibrahim Gokcen received his BS in Computer Engineering from Middle East Technical University in Ankara, Turkey in 1999. The same year, he joined the Department of EECS at Tulane University, where he currently is a research assistant in the Image Understanding and Evolutionary Computation lab and a PhD candidate. His research interests include statistical learning, pattern recognition and evolvable hardware.

Location: SEM 234

Monday, March 15th, 2004 at 4:00p.m.

Sponsored and organized by the CSE Seminar

Search Space Reduction Methods for Active Learning and Image Matching

In this talk, we formulate two different problems in statistical learning and image processing as search problems and present search space reduction methods for both.

Supervised learning requires prior labels for all training samples in order to find the decision boundary based on a given criterion of optimality. However, especially in the recent years, there has been an increase of applications generating huge amounts of data, majority of which is unlabeled. In the first part of the talk, we focus on the problem of active learning, which essentially aims at reducing the labeling cost by choosing unlabeled samples intelligently. This choice is made based on geometric, statistical and probabilistic measures employed on two isomorphic vector spaces and their subspaces. Two convex feasible regions of solutions are identified in both the parameter space (space of classifiers) and the input/feature space (space of samples). Estimation of the centroid or the volume of the two regions provide a way to assess search space reduction, thereby facilitating intelligent selection. Two practical applications of active learning on data thinning and image retrieval with relevance feedback are presented.

In the second part of the talk we present a rigid, search space reducing, feature based and adaptive image matching scheme to put images in correspondence without establishing explicit point correspondences. The method estimates the transformation parameters using a feature set, which is constructed via Principal Component Analysis (PCA). A unique aspect of the method is the incorporation of a learning process to learn the parameters incrementally from a training set of images. A nearest-mean matching scheme is used to match the features with transformation parameters. Hence, correct matching is determined within a predetermined error range.

Dr. Sergiu Dascalu

Assistant Professor
University of Nevada, Reno

Biography

Sergiu Dascalu is an Assistant Professor in the Department of Computer Science & Engineering at the University of Nevada, Reno. He has a Master's degree in Automatic Control and Computers from the Polytechnic Institute of Bucharest, Romania (1982), and a Ph.D. degree in Computer Science from Dalhousie University, Halifax, Nova Scotia, Canada (2001). Sergiu has been involved in the engineering of software for more than 15 years. He believes computers are essentially a fine combination of mathematics and arts, and endeavors to explore this combination with an engineer's discipline and dedication. His main research interests are in software engineering, particularly in languages, techniques, and tools for software specification. He has also worked on topics in the areas of real-time systems and human-computer interaction.

Location: MS 321

Thursday, March 11th, 2004 at 3:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Strata-Based Software Construction

Stratified Programming (SP) and its Strata-based Software Construction (SSC) extension are novel software development techniques proposed recently by Adrian Pasculescu, from Alpas Solutions, Toronto and Sergiu Dascalu, from the University of Nevada, Reno. In essence, by using strata to organize computer programs, both the design (software model) and the code (program implementation) can be adjusted to various levels of abstraction. Thus, a high level of flexibility is available both to designers who can work with the model at various levels of detail, and to users who can run various configurations of the same program (corresponding to various strata sets) much in the same way a computer game can be played at various skill-levels.

In essence, the strata-based approach allows enabling and disabling functionality and features of a program at both design stage and run-time. In this talk, the main stratified programming concepts are introduced, an example of SP code is shown, the use of strata in various phases of the software development process is discussed, and a number of directions for future work are presented.

Dr. Vanda Grubisic, Dr. Shulan Liu, and Mr. Mark Ballew

Desert Research Institute, Reno, NV

Biography

Dr. Vanda Grubisic - ACES Project Lead - received her Ph.D. in Atmospheric Science from Yale University in 1995. She is an Assistant Research Professor in the Division of Atmospheric Sciences at the Desert Research Institute with a research program in mesoscale meteorology, and the lead of the UCCSN Advanced Computing in Environmental Sciences (ACES) project. She came to DRI in 1999 from the National Center for Atmospheric Research (NCAR, Boulder, CO). Throughout her career, at both NCAR and DRI, Dr. Grubisic has been involved in high-performance computing projects with applications in atmospheric science. Her research focuses on numerical modeling and theoretical and observational analysis of mesoscale atmospheric flows, in particular airflow in complex terrain. She has extensive experience in development and use of atmospheric numerical models. She has used numerical models and advanced scientific visualization of the model data to investigate physical processes of gravity wave interaction with their critical levels, formation and stability of mountain wakes, evolution of Hawaiian rainbands, momentum transport by clouds in high-latitude cold-air outbreaks, and quantitative precipitation forecasting.

Dr. Shulan Liu - ACES Scientific Applications Programmer - obtained a BS degree in Physics from University of Science and Technology of China in 1997, and a Ph.D degree in computational physics from University of Massachusetts at Amherst in 2003. His Ph.D thesis was on the computer modeling of complex fluid. He joined the ACES program as a scientific application programmer in October 2003. Since then, he has been working on various aspects of the ACES program. In particular, he has been working on developing the ACES grid portal, creating the ACES grid, and porting simulation software to supercomputers.

Mr. Mark Ballew - ACES UNIX System Administrator - is currently working as a UNIX System administrator in the ACES program at the Desert Research Institute in Reno, Nevada to deploy a state wide data grid using Globus, data visualization technology, and Access Grid. With the Globus Toolkit, he is working to integrate idle computational resources state wide to distribute scientific workloads more efficiently. Using cutting edge hardware and software, he is also working on ways to visualize scientific data output in the ACES VisLab. In 2000 he graduated with a BS in Computer Science from the University of Nevada, Reno, and is currently pursuing a Master's degree in Computer Science from the same university.

Location: SLH 2

Thursday, February 26th, 2004 at 2:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Advanced Computing in Environmental Sciences (ACES): News and Review

The Advanced Computing in Environmental Sciences (ACES) program's main goals are creation of new capabilities for multidisciplinary research in Nevada, centered on computer modeling, data visualization, and other data-intensive techniques in environmental research, and building a critical mass of expertise and equipment to create a core for future computational science advances beyond the environmental applications. In this talk, we will review recent ACES developments, including new hardware at the ACES core computational facility at DRI, construction of the ACES VisLab at DRI, development of a new communication grid in the state with AccessGrid nodes, new ACES portal, and availability of technical support for computational science projects.

Dr. Andrew Knyazev

Associate Professor
University of Colorado at Denver, Department of Mathematics

Biography

Andrew Knyazev has received an MS degree in Computer Sciences from the Moscow State University in 1981 and a Ph.D. in Numerical Mathematics from the Institute of Numerical Mathematics Russian Academy of Sciences in 1985. After being employed as a Software Engineer at the Kurchatov's Institute of Atomic Energy in 1981-83, Andrew Knyazev started his research carrier at the Institute of Numerical Mathematics Russian Academy of Sciences, where he worked for nine years. In 1992, Andrew Knyazev has moved to the USA and became a visiting scientist at the Courant Institute of Mathematical Sciences, New York University for two years. Since 1994, he has been an associate professor at the Department of Mathematics, University of Colorado at Denver. One of the main research topics of Andrew Knyazev is numerical solution of large-scale symmetric eigenvalue problems on highly parallel computer systems and applications of eigenproblems in engineering. He is one of the contributors to the Templates for the Solution of Algebraic Eigenvalue Problems: A Practical Guide. Editors: Zhaojun Bai, James Demmel, Jack Dongarra, Axel Ruhe, and Henk Van der Vorst, SIAM, pp. 337-368, 2000.

Location: MS 321

Friday, February 20th, 2004 at 1:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Modern Preconditioned Eigensolvers for Spectral Image Segmentation

Several computationally efficient approaches to clustering large-scale and high-dimensional data are based on finding several extreme eigenvalues and corresponding eigenvectors of large sparse symmetric matrices. For example, a well-known problem of image segmentation can be mathematically formulated as a problem of minimization of normalized cuts, which leads to an eigenvalue problem.

Numerical solution of eigenvalue problems poses major computational challenges for large size matrices. As even low-end digital cameras produce mega-pixel images nowadays, it is typical to face image segmentation problems that lead to eigenvalue problems with millions unknowns. Eigenvalue problems for matrices so large are difficult to solve numerically at present. Moreover, it might be expected that the demands for the increased resolution and the needs to treat live feed dynamic images will continue to overgrow the increase in the computational power. An efficient choice of a method for numerical solution of the eigenvalue problem becomes crucial. The ultimate goal could be to find a method with a linear complexity, in other words, a method with computational costs that scale linearly with the problem size.

In the talk, we present an overview of numerical eigenvalue solvers used for image segmentation and introduce a locally optimal block preconditioned conjugate gradient (LOBPCG) method recently suggested and studied by the speaker. The LOBPCG is publicly available in MATLAB and in C using MPI and HYPRE libraries for massively parallel computers.

Dr. Sushil Louis

Associate Professor
University of Nevada, Reno

Biography

Dr. Louis is an associate professor of Computer Science & Engineering at the University of Nevada, Reno. He directs the evolutionary computing systems laboratory and works in the areas of evolutionary computation and machine learning. His current interests are in human behavior modeling and context aware computing. Major application areas within computer science include human computer interfaces, computer gaming, and computer security. More broadly, this work has applications in science, engineering, and business. You can visit his website at http://www.cse.unr.edu/~sushil/ for more information.

Location: SEM 234

Friday, February 13th, 2004 at 3:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Evolutionary Computing in Design

Natural selection has, over time, produced a variety of good designs and can be viewed as a search for higher fitness structures that perform specific functions. Evolutionary computing, also based on the principles of natural selection, views solving design problems as searching for solutions that meet specifications in a space of possible candidate solutions. This talk introduces evolutionary computing in the context of a simple problem and presents an overview of evolutionary computing in design. I use examples from designing combinational logic cicuits, trusses, finding shortest length tours, predicting subsurface structure, and asset allocation to illustrate this view. A short description of research at the Evolutionary Computing System Laboratory at UNR will conclude the talk.

Ian Parmee

University of the West of England, Bristol

Biography

Professor Ian Parmee has several years experience in both the Contracting and Consultancy sectors of the engineering industry. He returned to an academic career in 1991 and played a major role in the establishment and subsequent development of the Plymouth EPSRC Engineering Design Centre at the University of Plymouth investigating the integration of Evolutionary Computing technologies with Engineering Design. Ian was Director of the Centre from 1996 to 2000. He has now taken up a Professorship at the University of the West of England, Bristol where he is currently establishing research in the area of computational intelligence in design and decision-making through the ACDDM Group. His research has resulted in over one hundred publications in journals, conference proceedings and books. His recent book Evolutionary and Adaptive Computing in Engineering Design (Springer Verlag, 2001) describes much of his previous work. In addition, Ian chairs the International Conference on Adaptive Computing in Design and Manufacture which next takes place in April, 2004.

Location: SEM 326

Monday, January 26th, 2004 at 4:00p.m.

Sponsored and organized by the CSE / IEEE / EE Seminar

Handling Uncertainty and Poor Definition through Human-centred Evolutionary Computing and Complementary Agent-based Systems

Ill-definition, uncertainty and multiple objectives are primary characteristics of real-world design and decision-making processes. During the initial stages of such processes little knowledge appertaining to the problem at hand may be available. A primary task relates to improving problem definition in terms of variables, constraints and both quantitative and qualitative objectives. The problem space tends to develop with information gained in a dynamical process where optimization plays a secondary role following the establishment of a well-defined problem domain.

The talk presents an overall strategy involving evolutionary computing and complementary computational intelligence techniques within a human-centred, highly interactive design / decision-making environment. The main aim of this environment is the generation and succinct presentation of information relating to relationships and dependencies within complex multi-variate problem spaces. The intention is that the presentation of differing perspectives relating to complex interactions will contribute to a better overall understanding of prime problem characteristics and subsequent improvement in problem representation and solution identification.