CS Seminars

June 7, 2013 -- 1:30-2:30 PM in CCSB G.0208

Vilem Novak, Irina Perfilieva, University of Ostrava

Title: Analysis and Forecasting of Time Series using Fuzzy Transform and Fuzzy Natural Logic

Affiliation:
Centre of Excellence IT4Innovations, division of the University of Ostrava,
Institute for Research and Applications of Fuzzy Modeling,
Ostrava 1, Czech Republic

Abstract: A new methodology for analysis and forecasting of time series is introduced. It directly employs two non-statistical techniques: fuzzy transform (F-transform) and methods of fuzzy natural logic (FNL). Due to their use and the approach consisting in construction of several independent models, the methodology is capable at robust long time predictions.

Our methodology employs the decomposition method, namely, the time series is decomposed into 3 components: trend-cycle, seasonal component, and random disturbances. Unlike traditional approach, which assumes the trend-cycle to be an a priori given function for whole domain, we propose the use of fuzzy transform. Hence, the shape of the trend-cycle is not fixed to any specific function. It has been proved that the F-transform can eliminate seasonal component and significantly reduce noise. Moreover, the computational complexity is low.

For the trend forecast we employ the perception-based logical deduction (PbLD) applied to automatically generated fuzzy/linguistic rules. These methods were developed in FNL. The generated linguistic description characterizes future course of the trend-cycle. We can easily interpret it because it is formulated in (a fragment of) natural language. The seasonal component is analyzed and forecasted using combination of techniques which include again the fuzzy transform.

The lecture will be accompanied by demonstration of the experimental software system LFL Forecaster whose forecasting precision is fully comparable with professional forecasting systems such as ForecastPro (TM).

May 3, 2013 -- 1:00-2:00 PM in CCSB G.0208

Nigel Ward, UTEP

Title: Towards Dynamic Control of Voice Code Datarate using Importance Estimation

Abstract: Every cell phone and every Skype connection relies on a voice coder-decoder to digitize the signal and then reconstruct it. For each codec, the bitrate, and thus the signal quality, is typically constant over time. However some things people say are important, and some less so, meaning that a codec might dynamically allocate the bitrate (and thus the quality) depending on importance, and thereby be both more efficient and higher in perceived quality. We sketch a design for such a codec, and, focusing on the challenge of predicting importance in real time, report some initial positive results using prosodic features.

April 26, 2013 -- 1:00-2:00 PM in CCSB G.0208

Md. Jahid, Adrian Veliz, Rogelio Ochoa Jr., Gabriel R. Arellano, David Pruitt, Ed Hudgeons, and Eric Freudenthal, UTEP Robust Systems Group

Title: Garbage Collection Scheduling Strategies for Low Power, Memory Limited, and Delay Intolerant Devices (e.g. Smartphones)

Abstract: Languages that require automatic memory management are becoming increasingly dominant in a variety of delay-intolerant domains, including mobile devices and Web services. Automatic memory management requires periodic garbage collection, a computation that consumes substantial time and energy. The duration, frequency of, and efficiency of garbage collection depend strongly on heap size, which may be limited by available memory. Researchers have looked at the design of heuristics to control heap size and the scheduling of garbage collection to minimize energy consumption, computation time or delays but have paid little attention to the increasingly important combination of energy consumption, pause time, and memory size.

Our work examines this problem in the context of the Dalvik (J)VM within Google's popular Android mobile device platform. We observe that the heuristics employed by Dalvik lead it to enter degenerate configurations while executing interactive apps in which much power is consumed by repeated low-yield garbage collection operations that delay execution. We developed alternative heuristics to manage the scheduling of garbage collection scheduling and control heap growth, plus an instrumentation framework to support their evaluation. Initial results are encouraging. In this presentation, we describe the heuristics implemented by Dalvik and how they lead to the degenerate behaviors we observed. We also describe our alternative heuristics and instrumentation framework, results from initial experiments, and planned follow-on work.

April 19, 2013 -- 10:30-11:15 in Quinn Hall 212

Daniel A. Jimenez, Texas A&M University

Title: Reducing Wasted Speculation

Abstract: Modern microprocessors achieve high performance through aggressive speculation. However, large amounts of energy and potential performance are lost by speculating fruitlessly. The two most important speculation techniques are caches and speculative execution.

Caches hold a subset of the blocks from the high-latency main memory, speculating that quick access to these blocks will benefit the program. Unfortunately, most blocks in the last-level cache will not be referenced again before they are removed from the cache. These dead blocks waste time and energy as they reduce the effective capacity of the cache.

Speculative execution mitigates pipeline control hazards by predicting the outcome of branches, allowing subsequent instructions to be fetched and executed down the predicted path. Many instructions will be wrongly executed before an incorrect prediction is discovered, again wasting time and energy.

This talk discusses novel techniques for reclaiming lost performance and energy through reducing speculation wasted by caches and speculative execution. The talk will also discuss ongoing projects and future research directions.

Speaker Bio: Daniel A. Jimenez is an Associate Professor in the Department of Computer Science and Engineering at Texas A&M University. Previously he was a full Professor and Chair of the Department of Computer Science at The University of Texas at San Antonio and Associate Professor (with tenure) in the Department of Computer Science at Rutgers University. His research focuses on microarchitecture and low-level compiler optimizations. He introduced and developed the perceptron branch predictor which has inspired the design of two implemented microarchitectures: the AMD "Bobcat" core and the Oracle SPARC T4. Daniel earned his B.S. (1992) and M.S. (1994) in Computer Science at The University of Texas at San Antonio and his Ph.D. (2002) in Computer Sciences at The University of Texas at Austin. From 2002 through 2007, Daniel was an Assistant Professor in the Department of Computer Science at Rutgers. In 2005 Daniel took sabbatical leave at the Technical University of Catalonia (UPC) in Barcelona, Catalonia, Spain. In 2008 he was promoted to Associate Professor with tenure at Rutgers. He returned to his native Texas to take a position at UT San Antonio. He recently returned from a second sabbatical leave in Spain at the Barcelona Supercomputing Center and was General Chair of the 2011 IEEE HPCA conference. He is an NSF CAREER award recipient and ACM Distinguished Scientist.

April 19, 2013 -- 11:15-12:00 in Quinn Hall 212

Cynthia Phillips, Sandia National Laboratories

Title: Sensor Placement For Municipal Water Networks

Abstract: We consider the problem of placing a limited number of sensors in a municipal water distribution network to minimize the expected impact over a given suite of contamination incidents. In its simplest form, the sensor placement problem is a p-median problem that has structure extremely amenable to exact and heuristic solution methods. We describe the solution of real-world instances using integer programming (IP) and local neighborhood search. In this setting, the heuristic dominates integer programming in practice.

For versions with side-constraints, one can consider heuristics based on rounding a linear-programming (LP) relaxation. We will discuss two ways to round fractional solutions subject to a single cardinality constraint. Each has some provably nice properties with respect to the probabilities given by the LP relaxation.

When placing sensors to minimize the worst case impact over a set of contamination scenarios, the natural IP formulation is extremely difficult to solve. However, a reformulation based on iterated set cover is more accurate and much faster than the corresponding version of neighborhood search. This is particularly useful for finding solutions that give excellent tradeoffs between average-case and worst-case impacts.

Finally, we can take advantage of the practical solvability of the p-median-based version of the problem to approximately solve more realistic non-linear formulations such as a version that partially accounts for sensor failures.These algorithms are or will be incorporated into the TEVA-SPOT toolkit, a software suite that the US Environmental Protection Agency has used and is using to design contamination warning systems for US municipal water systems.

This is joint work with many colleagues at the EPA and at Sandia National Laboratories. Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-AC04-94AL85000.

Speaker Bio: Cynthia Phillips is a senior scientist in the Discrete Mathematics & Complex Systems Department at Sandia National Laboratories. She received a PhD in computer science from MIT in 1990. In her 23 years at Sandia National Laboratories she has conducted research in combinatorial optimization, algorithm design and analysis, and parallel computation with applications to many areas including scheduling, network and infrastructure surety, wireless network management, integer programming, cybersecurity, graph algorithms, and streaming algorithms.

April 19, 2013 -- 1:00-2:00 in CCSB G.0208

Daniel A. Jimenez and Cynthia Phillips

Title: Panel: Opportunites in Computer Science

Abstract: This panel discussion will focus on opportunities for graduate studies and research opportunities in computer science. It will include discussion of career paths and opportunities in both industry and academia, and will be targeted at undergraduate students interested in pursuing advanced studies in computer science or electrical engineering.

April 16, 2013 -- 2:00-3:00 PM in CCSB 3.0908 (Conference Room)

Maria Eskevich, Dublin City University

Title: Focus on Spoken Content in Multimedia Retrieval

Abstract: The volume of archived digital multimedia is growing rapidly with the expansion of technologies available for recording and storage. This material varies in topic, quality, source and form. While systems have been developed for browsing this data, less attention has been devoted to retrieval. Realizing the potential of this content requires users to be able to rapidly locate material relevant to their needs. While existing research has focused largely on search based on visual features for content where the relevant information is in the visual data stream, less attention has focused on data where the relevant information is predominantly in the spoken audio content.

This talk provides an overview of spoken content retrieval, ranging from formally structured broadcast news and movies to more informal content in the form of meetings, interviews, and semi-professional recordings on the Internet. As increasing multimedia usage is bringing new scenarios for spoken content retrieval, we explore potential directions for technology development through benchmarking initiatives, e.g. MediaEval.

In this talk we describe our work on test retrieval collection creation using crowdsourcing technologies, and we introduce our research into the development of speech search for challenging informally structured spoken content archives. We show in particular that different segmentation methods affect the result in the same way across different collection types and languages, and we explain the properties that a segmentation should have to best support spoken content retrieval.

Speaker Bio: Maria Eskevich is 4th year PhD Student in the School of Computing, Dublin City University, Ireland, working under the supervision of Dr. Gareth J.F. Jones. She holds a MSc. in Theoretical and Applied Linguistics from Saint Petersburg State University (Russia). She is one of the organisers of the Speech Retrieval, Search and Hyperlinking tasks at MediaEval Benchmark.

April 12, 2013 -- 1:00-2:00 PM in CCSB G.0208

Guillaume Melquiond, Universite Paris Sud

Title: Formal proof and interval arithmetic, a virtuous circle

Abstract: During the last fifty years, interval arithmetic has evolved continuously and methods for efficiently computing tight enclosures of expressions have improved. Despite these evolutions, reliability has always been preserved: whatever the round-off errors and the model errors, the computed enclosure is guaranteed to contain the mathematical result.

During this same period, formal systems have grown increasingly powerful. Their purpose is to mechanically verify mathematical theorems by checking every deduction step of their proofs. This ensures the highest confidence in proved properties, which might be mandatory for designing critical systems. The main drawback of formal systems, however, is that they rely on overly detailed proofs, and their use is thus time-consuming. This explains why proof assistants put a great emphasis on automation.

Thanks to its reliability, interval arithmetic seems like a good way to automate formal proofs, especially when it comes to properties about real numbers. Unfortunately, the strength of formal systems is that they never compromise with safety: since numerical computations performed by an interval package do not reduce to elementary deduction steps, they cannot be checked and thus trusted by formal systems. The challenge lies in reconciling both approaches.

In this talk, I will present how formal proofs and interval arithmetic interact in the context of the Coq proof assistant. In particular, I will show what it means to formalize interval arithmetic and how one can perform floating-point computations inside a formal system to prove mathematical theorems. The talk will be illustrated by various examples from the field of formal verification of numerical programs.

April 5, 2013 -- 1:00-2:00 PM in CCSB G.0208

William Yeoh, NMSU

Title: Distributed Constraint Optimization: Model, Algorithms, and Applications

Abstract: A Distributed Constraint Optimization Problem (DCOP) is a problem where several agents coordinate with each other to take on values so as to minimize the sum of the resulting constraint costs, which are dependent on the values of the agents. DCOPs are rapidly becoming popular for formulating and solving multiagent coordination problems such as the distributed coordination of sensors in a network and the distributed scheduling of meetings. Since solving DCOPs optimally is NP-hard, solving large problems efficiently becomes an issue. Nonetheless, researchers have recently developed a class of DCOP algorithms that use search techniques to optimally solve them. For example, ADOPT (Asynchronous Distributed Constraint Optimization) is one of the pioneering DCOP search algorithms that is widely extended.

In this talk, I will first describe the formulation for a DCOP and the motivation for using DCOPs in multiagent coordination problems. I will then describe Branch-and-Bound ADOPT (BnB-ADOPT), a memory-bounded asynchronous DCOP algorithm that uses the message passing and communication framework of ADOPT but changes the search strategy of ADOPT from best-first search to depth-first branch-and-bound search. Our experimental results show that BnB-ADOPT is up to one order of magnitude faster than ADOPT. Lastly, I will wrap up the talk by briefly describing our recent work on applying DCOP algorithms to solve smart grid optimization problems.

Bio: William Yeoh is an assistant professor of computer science at New Mexico State University. He received his Ph.D. in computer science at the University of Southern California. His research interests include multi-agent systems, distributed constraint reasoning, heuristic search, and planning with uncertainty. He was a coorganizer of the 2008 and 2013 International Workshop on Distributed Constraint Reasoning and the 2011 AAAI Symposium on Multiagent Coordination under Uncertainty.

Thursday March 28, 2013 -- 1:00-2:00 PM
Cyber-ShARE Vizualization Lab, Classroom Building Rm 401

Eric Sills, NC State

Title: Development and Evolution of Virtual Computing Lab Services at NC State University

Abstract: Over the summer of 2004 in response to a variety of demands that were not being met or that were being met poorly a small group with members from two campus IT organizations cobbled together a number of existing tools and technologies with a customized web application to deliver applications to students' computers from university servers. The imperative to "keep it simple" so as to be able to deliver a working system for pilot course use during the 2004 Fall semester turns out to have produced a flexible system that has been expanded to deliver a range of services never envisioned during the initial development.

This talk will discuss the demands that led to the development of the Virtual Computing Lab and the evolution of the Virtual Computing Lab services since 2004 including improved scheduling, long term server reservations, and clusters on demand. Features currently under development and potential future enhancements and directions will also be discussed.

Biography: Eric Sills currently serves as Executive Director of Shared Services in the Office of Information Technology at North Carolina State University (NC State). Eric received his BS, MS, and PhD degrees in Nuclear Engineering from NC State in a previous century. Eric worked for thirteen years in the scientific support group at the North Carolina Supercomputing Center prior to its closing in 2003. Following the close of the supercomputing center Eric hitched a ride with a p690 that was being transferred to NC State. Back at NC State Eric assisted with the implementation of a campus high performance computing service and later with the development of a virtual computing lab service utilizing much of the HPC technology. In addition to HPC and VCL services, the Shared Services department includes Unix/Linux based middleware services such as database support, web and app server support, identity and access management, and production workload management.

Joint CS/ECE Seminar

Friday March 15, 2013 -- 1:00-2:00 PM in CCSB G.0208

Jeanine Cook, NMSU

Title: Using Monte Carlo Techniques to Speed the Multicore Design Cycle

Abstract: Cycle-accurate simulation is the dominant methodology for processor design space analysis and performance prediction. However, with multi-core multi-threaded architectures, this method has become highly impractical as the sole means for design due to its extreme slowdowns. A statistical technique for modeling single- and multi-core processors that is based on Monte Carlo methods will be presented. We show that processor models of contemporary architectures can be developed and applied to performance prediction, bottleneck detection, and design space analysis.

Biography: Jeanine Cook is currently an Associate Professor in the Klipsch School of Electrical and Computer Engineering at New Mexico State University. She is the director of the Advanced Computer Architecture Performance and Simulation (ACAPS) Laboratory in the Klipsch School, where she and her students conduct research in the areas of processor and system simulation techniques, power optimization, performance modeling and prediction, workload characterization, and HPC emerging architectures. She received the B.S. degree in Electrical Engineering from the University of Colorado, Colorado Springs, in 1987, the M.S. degree in Computer Science from the University of Colorado, Boulder, in 1996, and the Ph.D. degree from New Mexico State University in 2002. She is a recipient of the 2008 Presidential Early Career Award (PECASE) for her research in computer processor performance modeling and she received the Frank Bromilow Excellence in Research award in 2009.

Interdisciplinary Symposium on Decision-Making Research

Thursday February 14, 2013 -- 9:00 AM - Noon

El Paso Natural Gas Conference Center

Valarie Reyna, Cornell University

James Stahl, Harvard Medical School

Title: Decision Analysis, Risk-Taking Behavior, and Healthcare

Valarie Reyna Bio: Valerie Reyna is Professor of Human Development and Psychology at Cornell University, Co-Director of the Cornell University Magnetic Resonance Imaging Facility, and a Co-director of the Center for Behavioral Economics and Decision Research. Dr. Reyna holds a Ph.D. in experimental psychology from Rockefeller University. Her research encompasses human judgment and decision making, numeracy and quantitative reasoning, risk and uncertainty, medical decision making, social judgment, and false memory. She is a developer of fuzzy-trace theory, a model of the relation between mental representations and decision making that has been widely applied in law, medicine, and public health. Her recent work has focused on aging, neurocognitive impairment, and genetic risk factors (e.g., in Alzheimer's disease); rationality and risky decision making, particularly risk taking in adolescence; and neuroimaging models of framing and decision making. She has also extended fuzzy-trace theory to risk perception, numeracy, and dual processes in medical decision making by both physicians and patients.

James Stahl Bio: James Stahl is a board certified internist and practicing clinician. His work focuses on operations research, decision analysis, outcomes research, and industrial design and he has particular expertise in simulation modeling as applied to healthcare.Dr. Stahl attended medical school at McGill University and completed a joint internal medicine residency between NSUH - Cornell Medical School and Memorial Sloan Kettering Cancer Center. He then completed a National Library of Medicine fellowship in Clinical Decision Making, Informatics and Telemedicine at New England Medical Center - Tufts University School of Medicine. While there, in addition to his training he worked on the clinical decision analysis consult service, developed and helped implement several internal hospital guidelines and decision support tools and helped develop and grow the international telemedicine program. He then went to University of Pittsburgh, where he worked in the area of liver allocation policy and completed his MPH at the Graduate School of Public Health and received the award for best health policy thesis.

His research interests include discrete-event simulation, operations research, decision analysis, meta-analysis, cost-effectiveness analysis, utility assessment, game theory, market design, ethics in the context of limited medical resources and applying industrial design to problem solving in health care. His current primary areas of research are health care system redesign and organ allocation policy and the role of new organ replacement technologies.

Monday December 3, 2012 -- 2:00-3:00 PM in CCSB G.0208

Graciela Gonzalez, Arizona State University

Title: Mining Social Network Postings for Mentions of Potential Adverse Drug Reactions

Abstract: Pre-market testing of drugs produces reasonably high quality information about the efficacy of the drug as a treatment for the condition for which it was approved, but gives a very incomplete picture of the drug's safety. It is only after a drug is marketed and used on a more widespread basis over longer periods of time that it is possible to identify other effects, such as rare but serious adverse effects, or those that are more common in the special subgroups excluded from the trial, among others. Post-marketing surveillance currently relies on voluntary reporting to the FDA by health care professionals (and recently, patients themselves).

Self-reported patient information captures a valuable perspective not captured by other means, and has been found to be of similar quality to that provided by health professionals. However, the value of numerous, informal self-reports such as those found in social network postings has not been evaluated. Through recently awarded NIH/NLM funding, we are deploying the infrastructure needed to explore the value of such postings as a source of "signals" of potential adverse drug reactions soon after the drugs hit the market. Despite the significant challenge of processing colloquial text, our prototype study showed promising performance, reaching an f-measure of 73.9% in identifying adverse reactions. Additional evaluation on un-annotated comments revealed encouraging correlations between adverse drug reactions found by our system and the documented reactions for those drugs. An overview of the methods and findings of this project will be discussed in this presentation.

Friday November 9, 2012 -- 1:30-2:30 PM in Biosciences Research Building Room 2.168

Dr. Michel Dumontier, Carleton University, Canada

Title: From Biological Data to Clinical Applications: Positioning a digital infrastructure for the future of biomedicine

Abstract: In the quest to translate the results of life science research into effective clinical applications, many are now turning their attention to and also trying to make sense of the large and rapidly growing amount of biological and biomedical data. Indeed, getting a grip on and keeping on top of the daily flood of new information, whether it be the latest in clinical reviews, scientific reports, or raw data is an ever-present and widely-recognized challenge. The limited access to structured, integrated and citable data limits our ability to exploit a rich source of scientific knowledge for clinical and translational research. While keeping the dual goals of increasing our understanding of how living systems respond to chemical agents and translating our combined knowledge into clinical applications, I will discuss our efforts to leverage Semantic Web technologies to facilitate the formulation, publication, integration, and discovery of biological facts, expert knowledge and services of value to pharmaceutical and clinical research, and more recently, with applications for the patient-centric delivery of health care.

Speaker's Bio: Dr. Michel Dumontier is an Associate Professor of Bioinformatics in the Department of Biology, the Institute of Biochemistry and School of Computer Science at Carleton University in Ottawa, Canada. His research aims to develop semantics-powered computational methods to increase our understanding of how living systems respond to chemical agents. At the core of the research program is the development and use of Semantic Web technologies to formally represent and reason about data and services so as (1) to facilitate the publishing, sharing and discovery of scientific knowledge produced by individuals and small collectives, (2) to enable the formulation and evaluation scientific hypotheses using our collective tools and knowledge and (3) to create and make available computational methods to investigate the structure, function and behaviour of living systems. Dr. Dumontier is the Scientific Director for the open-source Bio2RDF linked data for life sciences project and currently serves as a chair for the World Wide Web Consortium Semantic Web in Health Care and Life Sciences Interest Group (W3C HCLSIG). Dr. Dumontier is as an editor for the open-review Semantic Web Journal, co-chairs the ISMB-affiliated Bio-Ontologies Special Interest Group, and is a co-chair for the 2013 Semantic Trilogy consisting of the International Conference on Biomedical Ontology (ICBO), the Canadian Semantic Web Working Symposium (CSWWS) and Data Integration in the Life Sciences (DILS).

Friday October 26, 2012 -- 1:00-2:00 PM in CCSB G.0208

Shirley Moore, UTEP

Title: DOE SciDAC Climate Modeling Research

This talk will be accessible to undergraduate students.

Abstract: The Department of Energy has invested heavily in high resolution global climate simulation. The problem of predicting climate change and its consequences is motivated by the increasingly urgent need to adapt to near term trends in climate change and the potential changes in the frequency and intensity of extreme events. There is a growing need for scientific insights regarding climate change, its consequences, and how organizations ranging from local communities to national government and international agencies can best respond to challenges in both adaptation to and mitigation of climate change. This talk will discuss the scientific consensus on climate change--what we know, how we know it, and what we don't know and need to know.

Wednesday October 10, 2012 -- 1:30-3:00 PM in CCS G.0208

Dr. Jean-Charles Regin, University of Nice-Sophia Antipolis, France
Website: http://constraint-programming.com/people/regin/index.html

Title: Embedding OR into CP

Abstract: Constraint programming (CP) is a general and powerful method to solve some combinatorial problems. This method has been successfully used to solve a large range of real-life applications (rostering, time-tabling, car manufacturing, scheduling etc...), which can be quite different.

CP is on the borderline of AI and OR. CP inherits from AI the language aspect (that is the easy way to define a problem), the flexibility (that is the fact that problems can be easily modified) and the possibility to benefit from the knowledge of the application domain to improve the resolution of a problem. Operational Research brings to CP a lot of nice and powerful algorithms which gives to CP the computational aspect needed to solve some real world problems.

In this talk we will introduce the general principles of this method. We will show its strength and its weakness. Notably, we will insist on one major strength of CP: the Global Constraints. We will see how Global Constraints can be designed and we will give some general guidelines to create new Global Constraints. We will also detail how some OR algorithms are used within Global Constraints and we will present how it leads to new research areas. Some computational results will be given.

Keywords: Constraint Programming, Operation Research, Flow, Matching, Filtering Algorithms.
No prerequisite knowledge is required.

Friday September 28, 2012 -- 12:00-1:00 PM in CCS G.0208

Leonardo Salayandia, UTEP

Title: Ontologies for Scientific Data Transformation

Abstract: A common task for scientists is to collect data and apply algorithms to process the data. Scientists who aim to use data collected or processed by other scientists must be able to determine the applicability of the data. This requires associating the data with metadata about the conditions under which the data was collected and how it was processed, i.e., its provenance. This talk will present the Workflow-Driven Ontology (WDO) framework that supports the ability of scientists to formally represent knowledge about the processes used to collect data and transform it into data products. Formally represented knowledge supports computing tasks that require reasoning capabilities, allowing scientists to use computers to support decisions about the appropriateness of data for reuse. The approach uses abstraction from the perspective of scientists to identify relevant components in collecting and transforming data, postponing technical nuances to a later stage. In addition, the WDO framework is useful to guide the capture of data provenance at relevant levels of granularity for the scientist.

An implementation of this framework, called WDO-It!, uses Semantic Web technologies to encode formal knowledge representations, as well as the Proof Markup Language (PML) to capture data provenance. WDO-It! has been used in projects related to Environmental Sciences and Geosciences to document data collection and transformation processes, as well as to record provenance of data products. WDO-It! is also being adopted by projects related to the Atmospheric Radiation Measurement Program at Pacific Northwest National Laboratory.

Leonardo Salayandia is a research specialist at the NSF-funded CyberShARE Center at the University of Texas at El Paso. He is a Computer Science Ph.D. candidate expecting to complete his doctoral work in Fall 2012. His areas of interest include formal knowledge representation, software engineering, and computer architecture. He is also interested in Semantic Web technologies and their application to scientific data management.

Friday May 18, 2012 -- 1:30pm in CCS G.0208

Vladik Kreinovich, UTEP

Title: Infinitesimals: From Informal Calculus Ideas of Newton and Leibniz to Modern Consistent Theory, with Applications to Decision Making and Uncertainty Analysis

Abstract: Traditionally, in decision making, we assume that we know the objective function f(x) that describes the quality of different alternatives x, and out of all possible alternatives, we select a one for which the value f(x) is the largest (or the smallest).

Sometimes, we have several alternatives with the same value of the objective function f(x). For example, when we select a numerical method for solving a given problem (e.g., a system of differential equations), we usually want to minimize the approximation error of the resulting solution -- e.g., gauged by the mean squared approximation error. If we have two algorithms with the exact same value of y=f(x), we can use this non-uniqueness to optimize something else: e.g., select a method for which the average computation time z=g(x) is the smallest possible. The resulting preference relation < corresponds to lexicographic order: an algorithm x with (y,z)=(f(x),g(x)) is better than an algorithm x' with (y',z')=(f(x'),g(x')) if either y < y' or ( y=y' and z < z').

With respect to this relation, the pair (0,1) is better than (0,0), but worse than (1/n,0) for all n > 0. Such objects -- which are larger than 0 but smaller than any positive real numbers -- are known as "infinitesimals". Infinitesimals dx, dt, etc., were used by Newton and Leibnitz in their original formulation of calculus. The original treatment of infinitesimals was prone to paradoxes and conflicting results until, in the 19 century, Cauchy and Weierstrass came up with the more familiar epsilon-delta definitions. Infinitesimals were revived in the 1960s by a logician Abraham Robinson who provided a consistent way of treating infinitesimals (and their reverses -- true infinite values) now known as non-standard analysis.

Infinitesimals are also used in interval computations (and, more generally, in uncertainty analysis) -- details will be explained in the talk. (And infinities also appear in quantum field theory.)

There are many different schemes for introducing infinitesimals. Somewhat surprisingly, they all turn out to be particular cases of a general scheme of surreal numbers originally proposed by J.H.Conway. This general scheme will be described in this talk.

The talk is largely based on a recent paper P. Erlich, "The absolute arithmetic continuum and the unification of all numbers great and small", The Bulletin of Symbolic Logic, 2012, Vol. 18, No. 1, pp. 1-45.

Friday April 13, 2012 -- 1:30pm in CCS G.0208

Irina Perfilieva, University of Ostrava, Czech Republic

Title: F-transform in Computer Vision

Abstract: The theory of F-transform (short name for fuzzy transform) is a modern theoretical tool for fuzzy modeling. On the basis of this theory, a methodology with many applications in the areas of data analysis, image processing, time series analysis and forecasting has been developed. Due to clear theoretical basis and common effort of many researches, both theory and applications have been extensively developed in recent years.

The theory of F-transform has a comprehensive mathematical background which is comparable with the well-known transforms such as Laplace, Fourier, and wavelet ones. Its results are comparable and in some respects surpass the classical models, for example in time series analysis and forecasting, in the analysis of economic data, or in image processing.

In the talk we plan the following:

1. introduce backgrounds of the F-transform methodology including simple and attractive properties of its direct and inverse parts,
2. show and explain successful applications of the F-transform methodology in image processing; namely, edge detection, compression, reconstruction, fusion.
3. show real-life applications in non-invasive quality control and image recognition.

The presentation will proceed on a rather elementary level with many graphical illustrations. No special knowledge in mathematics is required.

Friday, April 6, 2012 -- 1:30pm in CCS G.0208

Branislav Bosansky, Czech Technical University in Prague

Title:Game Theory for Patrolling Games with Mobile Players

Abstract: Game theory provides theoretic and algorithmic foundations for the field of multi-agent systems by formal modeling of situations with non-cooperative self-interested autonomous agents, and by defining the optimal behavior of the agents in such situations. The game-theoretic concepts can be used for design of software multi-agent systems, but also for intelligent applications that assist human decision making in various scenarios populated by adversarial agents. Examples can be found in computer networks connecting legitimate users and attackers, public transport system providing services to paying customers and fare-evaders, or decentralized surveillance system protecting buildings or utility infrastructure against thieves and terrorists alike.

In the talk we will describe security games with mobile players -- graph-based security games and patrolling games -- and we will discuss the way of applying the game-theoretic methods into a multi-agent simulation and verifying the proposed models and designed algorithms.

Bio: Branislav Bosansky is a PhD student working in Agent Technology Center at Department of Computer Science and Engineering at Czech Technical University in Prague. He holds master's degree in theoretic informatics (artificial intelligence and logic programming) from Faculty of Mathematics and Physics at Charles University in Prague. His research interests include computational game theory with applications in multi-agent environments, game-playing algorithms, and multi-agent simulations. He participated in a number of successful projects supported by US Air Force, Office for Naval Research, and CERDEC US Army. The focus of his work was on designing and implementation of novel algorithms for finding strategies in complex adversarial domains, such as humanitarian relief operation in an unstable region, problem of piracy in Gulf of Aden and Indian Ocean, and tactical military operations with teams of heterogeneous unmanned systems. The game-theoretic models of multi-stage security games and patrolling games are the topic of his thesis.

Tuesday, April 3, 2012 -- 2:00pm in CCS 3rd floor conference room

Branislav Bosansky, Czech Technical University in Prague

Title:Applications of Agent Technologies

Abstract: Agent Technology Center (ATG) is a research group at Department of Computer Science at Czech Technical University in Prague. Research in the group is mainly focused on multi-agent simulations, agent-based computing and optimization, and research contributions vary from theoretic topics such as computational game-theory or planning, through implementing and evaluating developed algorithms in high-fidelity multi-agent simulation, and finally to applying these methods in robotics and testing the algorithms in real world settings. The talk will introduce main long-term research tracks in ATG: group of projects focused on air-traffic simulation, execution of tactical missions by UAVs, and deployment of designed algorithms to real UAVs; securing maritime transit using multi-agent simulation and related projects concerned with methods of transportation in urban areas; and projects related to the computer-network security. Finally, we look at the applications of computational game theory in our projects and describe several use-cases, where the game-theoretic approach was successfully used.

Bio: Branislav Bosansky is a PhD student working in Agent Technology Center at Department of Computer Science and Engineering at Czech Technical University in Prague. He holds master's degree in theoretic informatics (artificial intelligence and logic programming) from Faculty of Mathematics and Physics at Charles University in Prague. His research interests include computational game theory with applications in multi-agent environments, game-playing algorithms, and multi-agent simulations. He participated in a number of successful projects supported by US Air Force, Office for Naval Research, and CERDEC US Army. The focus of his work was on designing and implementation of novel algorithms for finding strategies in complex adversarial domains, such as humanitarian relief operation in an unstable region, problem of piracy in Gulf of Aden and Indian Ocean, and tactical military operations with teams of heterogeneous unmanned systems. The game-theoretic models of multi-stage security games and patrolling games are the topic of his thesis.

Friday March 30, 2012 -- 1:30pm in CCS G.0208

Graciela Perera

Title: Monitoring Bridge Safety with Skyline Algorithms for Wireless Sensor Networks

Abstract: Physical and cyber infrastructure are the building blocks of many critical systems that protect and provide safety. Incidents like the collapse of the Interstate 35W bridge in Minneapolis during 2008, illustrate the importance of monitoring bridges for damage. Current advances in Structural Health Monitoring (SHM) Systems emphasized using fresh data collected from Wireless Sensor Networks (WSN). This creates a challenge in the deployment of the WSN given their limitations in battery life, processing, and storage capabilities. We propose using the SUN SPOTS WSN technology from Oracle to study how can the Skyline BNL algorithm be used to extend the battery life of the WSN without discarding data required in the damage detection process. The results of this research will potentially increase the bridge safety and longevity. In addition, we would like to present our preliminary ideas related to malicious traffic detection using the NetFPGA; an open-source networking platform for classroom and research experimentation.

Friday March 23, 2012 -- 1:30pm in CCS G.0208

Luc Longpre, Department of Computer Science, UTEP

Title: Bootsrapping trust in commodity computers

Abstract: How do you run some security-sensitive code on a computer where you believe that the operating may have been modified by sophisticated malware? We will survey different approaches to bootstrap trust in commodity computers. One such approach is flicker, an infrastructure for executing the security-sensitive code in isolation, with minimal assumptions. We will also present some recent attacks on such systems. This talk is a summary of 2 talks presented at UT Austin by different invited speakers: Bryan Parno and Hovav Shacham.

Monday March 19, 2012 -- 1:30pm in CCS G.0208

James Edmondson

Title: Algorithms, Mechanisms, and Tools for Enterprise Distributed Real-time and Embedded (DRE) Systems

Abstract: Enterprise distributed real-time and embedded (DRE) systems are an important class of applications whose stringent quality-of-service (QoS) requirements must be met despite operating in environments with finite CPU, memory, and network resources. In many cases, enterprise DRE systems are mission-critical and the failure of such systems to meet QoS requirements may result in catastrophic damage, financial loss, or even fatalities. Examples of enterprise systems include air traffic control, power grids, shipboard computing, fractionated satellite constellations, and search-and-rescue missions.

This presentation discusses algorithms, heuristics, and open-source middleware and tools that help to address key challenges in enterprise DRE systems, including (1) a distributed mutual exclusion algorithm called PADME that controls differentiated resource acquisition at-scale, (2) a high-performance knowledge and reasoning engine that infers important knowledge from a multitude of online sensors and actuators in real-time, (3) a deployment and testing engine for validating enterprise DRE systems prior to deployment in real-world environments, and (4) heuristics and algorithms for maintaining optimal deployments within a constrained environment via timely approximations of the subgraph isomorphic problem. The behavior and performance of these algorithms, heuristics, and open-source middleware and tools will be analyzed and future plans for tackling challenges in the DRE and cyber-physical systems areas will be presented.

Speaker Bio: James Edmondson is a doctoral candidate in the Department of Electrical Engineering and Computer Science at Vanderbilt University. He is also a graduate research assistant at the Vanderbilt University Institute for Software Integrated Systems, where he has been involved in providing high-performance solutions and testing infrastructure for enterprise distributed real-time and embedded (DRE) systems. His recent work focuses on the development of heuristics, algorithms, and technologies into generic open-source middleware and tools for use in highly constrained, large scale enterprise DRE systems.

Monday March 5, 2012 -- 2:00pm in CCS G.0208

Prasad Calyam

Title: User Quality-of-Experience Delivery through Context-aware Resource Adaptation

Abstract: With the increased prevalence of high-speed networks, end-users are relying on a variety of application services (e.g. telephony, videoconferencing, IPTV, data clouds) to be delivered via the Internet. The application services cater to end-users' needs that range from social communications/entertainment to virtual organizations/classrooms in science and engineering. However, the Internet was originally built on best-effort principles that make it susceptible to events such as multi-domain performance bottlenecks and cyber-attacks. Hence, there are serious challenges in delivering, securing and managing the Internet Quality of Service (QoS) levels in order to satisfy user Quality-of-Experience (QoE) expectations within applications. In my talk, I will describe our R and D efforts for improving user QoE through context-aware resource adaptation to deliver application services such as videoconferencing, IPTV and virtual desktop clouds in a scalable, reliable and secure manner. More specifically, I will present our innovations in application behavior modeling under different network-health and system-load contexts, and the use of these models in adaptations at 'end-user clients' as well as 'application service clouds'. I will conclude my talk with case studies of how our R and D outcomes are being adopted in "Future Internet" programs such as NSF GENI and DOE Next-generation Networking.

Speaker Bio: Prasad Calyam received his M.S. and Ph.D. degrees from the Department of Electrical and Computer Engineering at The Ohio State University in 2002 and 2007, respectively. He is currently a Senior Research Engineer at Ohio Supercomputer Center/OARnet, The Ohio State University. His research and development areas of interest include: Distributed Computing, Computer Networking, Cyber Security, and Cloud Computing. As the Principal Investigator, he has successfully led teams of graduate, undergraduate and postdoctoral fellows in numerous Federal, State and Industry sponsored research and development projects. Dr. Calyam's research sponsors include: DOE, NSF, VMware, Dell, IBM, Huawei Technologies, Apparent Networks, Polycom, Ohio Board of Regents, and Internet2. He has also given technical presentations on a regular basis at undergraduate/graduate classrooms as well as regional/national/international forums.

Wednesday February 29, 2012 -- 3:00pm in BIOS 2.168

Shahadat Hossain

Title: The CPR Method and Beyond

Abstract: In 1974, A. R. Curtis, M. J. D. Powell, and J. K. Reid published a seminal paper on the estimation of Jacobian matrices which was later coined as the CPR (Curtis-Powell-Reid) method. Central to the CPR method is the effective utilization of a priori known sparsity information: compressing the matrix, computing the nonzero unknowns using, for example, automatic differentiation (AD), and reconstructing the Jacobian. It is only recently that the optimal CPR method in its general form is characterized and the theoretical underpinning for the optimality is shown. In this talk we provide an overview of some of the recent developments in sparse Jacobian determination with a brief introduction to AD.

Monday February 27, 2012 -- 3:00pm in BIOS 2.168

Shirley Moore

Title: Forging Partnerships Between Computer and Computational Science: The Road Ahead

Abstract: High performance computing (HPC) environments and scientific application systems are in a period of rapid evolution and both are becoming increasingly complex. High performance architectures are evolving towards combining many processors with diverse architectures ranging from multicore chips through SIMD accelerators. These new and different types of architectures will require fundamental changes in algorithms and programming methods. Applications are adding additional science and pushing the envelope of scalability on current systems. The range of disciplines utilizing modeling and simulation is expanding and thus extending the range of types of computations executing on HPC systems. To succeed, application developers will need to reconsider both their code's structure and the tools they use to develop, tune, and run that code.

This talk will summarize recent collaborative work with application scientists in the areas of computational chemistry, quantum mechanics, and bioinformatics that is addressing the above challenges. Aspects of this work include implementation of hybrid programming models, performance evaluation and optimization, and I/O scaling. Next, the talk will describe current and planned work with scientists in the above computational science areas as well as in computational astrophysics. The current and future work involves refactoring and restructuring of long-lived large-scale applications; modeling, implementation, and optimization of algorithms on GPUs; and scaling of hybrid (MPI+threads) applications to hundreds of thousands of multicore nodes.

Finally, the talk will comment on the current landscape and future trends in funding opportunities for interdisciplinary computational science.

Wednesday February 1, 2012 -- 1:30pm in TBD

Teruhisa Misu, NICT

Title: Speech-based Interactive Information Guidance System Using Question-Answering Technique

Abstract: In this talk, a speech-based interactive guidance system based on document retrieval and presentation will be addressed. In conventional audio guidance systems, such as those deployed in museums, the information flow is one-way and the content is fixed. To make the guidance interactive, we prepare two modes, a user-initiative retrieval/QA mode (pull-mode) and a system-initiative recommendation mode (push-mode), and switch between them according to the user's state. In the user-initiative retrieval/QA mode, the user can ask questions about specific facts in the documents in addition to general queries. In the system-initiative recommendation mode, the system actively provides the information the user would be interested in. We implemented a navigation system containing sightseeing information. The effectiveness of the proposed techniques was confirmed through a field trial by a number of real novice users.

Teruhisa Misu received the B.E. degree in 2003, the M.E. degree in 2005, and the Ph.D. degree in 2008, all in information science, from Kyoto University, Kyoto, Japan. From 2005 to 2008, he was a Research Fellow (DC1) of the Japan Society for the Promotion of Science (JSPS). In 2008, he joined NICT Spoken Language Communication Group. His research interests include spoken language processing for spoken dialogue system, in particular dialogue modeling and management, question-answering, language modeling for speech recognition.

Monday January 30, 2012 -- 3:30 PM (Room TBD)

Malcom Gethers, Faculty Candidate

Title: An Inductive Framework to Support Software Maintenance

Abstract: Software maintenance and evolution is a particularly complex phenomenon in case of long-lived, large-scale systems. It is not uncommon for such systems to progress through years of development history, a number of developers, and a multitude of software artifacts including millions of lines of code. Therefore, realizing even the slightest change may not always be straightforward. Clearly, changes are the central force driving software evolution. Therefore, it is not surprising that a paramount effort has been (and should be) devoted in the software engineering community to systematically understanding, estimating, and managing changes to software artifacts. This effort includes the three core change related tasks of (1) concept or feature location - locating where a particular functionality is implemented in a code or a starting point of a change, (2) impact analysis/traceability link recovery - identifying which other software artifacts should be changed given an initial software artifact, and (3) expert developer recommendations - identifying who are the most experienced developers to implement needed changes. Our work defines a framework for an integrated approach to support three core software maintenance and evolution tasks: feature location, software change impact analysis, and expert developer recommendation. The approach is centered on the combinations of the conceptual and evolutionary relationships latent in structured and unstructured software artifacts. Information Retrieval (IR) and Mining Software Repositories (MSR) based techniques are used for analyzing and deriving these relationships. All the three tasks are supported under a single, common framework by providing systematic combinations of MSR and IR analyses on single and multiple versions of a software system. This combining ability sets it apart from previously reported relevant solutions in the literature.

Thursday January 26, 2012 -- 3:30 PM CCS Building G.0208

Collin McMillan, Faculty Candidate

Title: Searching, Selecting, and Synthesizing Source Code Components

Abstract: As programmers develop software, they instinctively sense that source code exists that could be reused if found -- many programming tasks are common to many software projects across different domains. Oftentimes, a programmer will attempt to create new software from this existing source code, such as third-party libraries or code from online repositories. Unfortunately, several major challenges make it difficult to locate the relevant source code and to reuse it. In this talk, I will discuss these challenges and our work in source code search and reuse. Our work relies on structural information such as function calls and Application Programming Interface (API) usage to locate relevant source code, which can be reused in programmers' projects.

Tuesday Jan. 24, 2012 -- 5:00 PM
Undergraduate Learning Center Rm 116

Part of the UTEP Centennial Lecture Series, hosted by President Diana Natalicio and the Department of Computer Science

Richard Talbot, Director, Product Line Management of the IBM Power Systems

Title:Beyond Jeopardy! Putting Watson to Work

Abstract: IBM Watson is an artificial intelligence system capable of answering questions posed in natural language-with precision, confidence and speed. Named after IBM's founder and first president, Thomas J. Watson, the supercomputer is one of the most advanced systems on Earth and was developed by an IBM Research team of 25 scientists. In an historic event, in February 2011, Watson competed on Jeopardy! against the TV quiz show's two most renowned champions, delivered a stellar performance and captured the world's imagination. The IBM Jeopardy! Challenge represented a major milestone in the development of artificial intelligence systems, and was highlighted in IBM's centennial celebration this year as a showcase for its commitment to research and tradition of Grand Challenges. While the Grand Challenge for this IBM Research team was winning the three game Jeopardy! tournament, IBM's bigger vision for Watson targets the development of broadly applicable, commercial technologies capable of digesting a variety of structured and unstructured information and responding with precise answers to questions posed in natural language. IBM has now challenged itself to successfully migrate Watson's sophisticated analytics capabilities to solving real world challenges in Healthcare, Finance, Banking and other industries. Richard Talbot, Director, IBM Power Systems will be presenting highlights of this exciting IBM Research program and delving into present and future business implications of the DeepQA and Natural Language Processing technologies behind Watson. After the presentation, a panel of business and research leaders from the El Paso community will discuss future implications of these exciting new technologies.

Biography: Tichard Talbot currently serves as Director, Product Line Management of the IBM Power Systems business. In this role, he leads development of portfolio definition, business planning and successful world-wide introduction of IBM's next generation Power platforms. His team contributes to the achievement of all Power Systems' business and quality objectives and helped grow this IBM business unit to #1 market share leadership.
Most recently, Richard has been working on the IBM leadership team responsible for transitioning Watson from the game show circuit to solving real world problems in healthcare, financial services, government citizen services and many other industries. He is personally involved in the development of several Watson pilot applications and works frequently with clients to understand how these new analytics and natural language processing technologies can help address their most complex business challenges and accelerate their delivery of new services and offerings. His interests also include developing broad scale IBM partnerships within the healthcare industry and university alliances for the deployment of these new technologies to accelerate the development of new treatment options and cures for complex disease processes.
Richard is an IBM Executive and PMI Certified Project Manager, holds five U.S. Patents and has received several awards in management excellence, business and technical achievement. Prior to this role, he held a number of management and software development roles since starting with IBM in Boca Raton, Florida.
Richard received his Master's degree in Electrical Engineering from the University of Texas in Austin and BA from Rice University. He currently resides with his family in Austin, Texas and remains active in several charitable and community service organizations.

Monday January 23, 2012 -- 10:00 AM in OLD CS Building Rm 321

Kristen Brent Venable, Faculty Candidate

Title: Sequential aggregation of compact preferences

Abstract: We consider scenarios where a set of agents needs to select a common decision from a set of possible decisions, over which they express their preferences. We also assume that such a decision set has a combinatorial structure, that is, that each decision can be seen as the combination of certain features, where each feature has a set of possible instances. This occurs in several AI applications, such as combinatorial auctions, web recommender systems, and configuration systems. Even if the number of features and instances is small, the number of possible decisions can be very large.Fortunately, in the presence of such a combinatorial structure, agents may describe their preference in a compact and efficient way, using one of the several formalisms available in the literature, such as, for example soft constraints and CP-nets. We consider a sequential procedure that chooses one candidate by asking the agents to vote on one feature at a time. We investigate this approach to preference aggregation from a computational social choice point of view. In particular, we study when and if desirable properties, such as, for example, anonymity, strategy-proofness and resistance to bribery hold.

Tuesday December 6, 2011 -- 2:00pm in CS 308 (NOTE special day and time)

Timo Baumann, Uni Hamburg

Title: Real-time End-to-end Incrementality in Spoken Dialog Systems

Abstract: Current spoken dialogue systems are not yet suitable for natural, conversational dialogue for a number of reasons. One of these is a lack of responsiveness. A way to overcome this shortcoming is incremental processing, that is, processing of user input while it is still ongoing. However, processing delays must be kept to a minimum in order to allow for advanced dialog behaviour such as giving feedback during a user's turn, interrupting, or co-completing the user.

As a proof of concept that end-to-end incremental dialog processing can work in real-time, I will present a system that is able to speak in synchrony with the user, that is, speak the same words at the same time as the user says them (given that the user's utterance is known).

Friday November 18, 2011 -- 1:30pm in CS 308

Nigel Ward, Department of Computer Science, UTEP

Title: Towards Responsiveness in Information-Exchange Dialogs

Abstract: Today's spoken dialog systems (voice user interfaces) are widely disliked for the bad user experiences they provide. At the same time, interaction by voice offers unique opportunities to provide comfortable and rewarding experiences. Indeed, one research group has demonstrated a dialog system augmented with responsive back-channeling that was perceived as a better listener than humans on average; other responsive behaviors relating to turn-taking and emotional dimensions are also effective.

We are about to start a new project, aiming to build such responsive behaviors into a dialog system for a real information-exchange domain. Challenges include: discovering responsive behaviors, which are often dependent on subtle aspects of tone of voice and prosody; coordinating these with the current dialog state, and concisely representing that state; adapting these behaviors to specific users and user types; and building a full system able to carry out a complete task with real users.

Friday November 4, 2011 -- 1:30pm in CS221

Vladik Kreinovich, Department of Computer Science, UTEP

Title: Decorated Intervals: A Way to Optimally Propagate Properties through Numerical Computations

Abstract: One of the main problems of interval computations is to find an enclosure Y that contains the range f(X1, ..., Xn) of a given function f(x1, ..., xn) over given intervals X1, ..., Xn. Most of the techniques for estimating this range are based on propagating the range through computations. Specifically, we follow the computations of f(x1, ..., xn) step-by-step: we start with ranges X1, ..., Xn of the inputs, and then we sequentially compute the enclosures for the ranges of all intermediate results, until, on the last computation step, we get the desired enclosure Y.
A similar propagation of "decorations" -- information about continuity -- enables us to make conclusions about the continuity of the resulting function f(x1, ..., xn). In this talk, we describe the corresponding algorithms, and explain how the interval propagation results can be naturally extended to the general case of arbitrary sets.

Friday October 7, 2011 -- 1:30pm in CS221

Deana Pennington, Cyber-ShARE Center of Excellence, UTEP

Title:Science and Technology Research Teams and the Fuzzy Front End of Innovation

Abstract: Science and technology has a long, intertwined history. At times science has driven technology and at other times technology has driven science. In recent years, funding agencies have frequently called for interdisciplinary research that is simultaneously innovative in both science and technology. That is, innovation is expected to co-emerge in multiple research areas based on synergistic interactions between diverse researchers. Though highly sought after, evidence of such outcomes is sparse. This seminar will provide an overview of our understanding of interdisciplinary research processes; how those processes can drive innovation; common barriers; and hypothesized mechanisms for overcoming barriers. In particular, theories of transformational learning will be invoked to explain the opportunities and challenges of confronting interdisciplinary research teams at their inception, during the fuzzy front end of innovation.

Friday September 23, 2011 -- 1:30pm in CS221

David Novick, Department of Computer Science, UTEP

Title:The Communicative Functions of Animation in User Interfaces

Abstract: To develop a model that relates the purpose of the communication to the nature of the animation, we surveyed existing user interfaces that use animation, analyzed these uses with respect to type of animation and communicative function, and considered ambiguous or otherwise difficult cases. From this analysis, we constructed a matrix with appropriateness/inappropriateness values for all combinations of communicative functions and animation types covered by our survey. To illustrate how the model could be applied to graphical user interfaces and to assess the model's plausibility, we used the model to develop two versions of a user interface for an MP3 player.

Friday September 9, 2011 -- 1pm in CS221

Eric Freudenthal, Department of Computer Science, UTEP

Title: Planting the seeds of computational thinking: An introduction to programming suitable for inclusion in STEM curricula

Abstract: Inadequate math preparation discourages many capable students - especially those from traditionally underrepresented groups - from pursuing or succeeding in STEM academic programs. iMPaCT is a family of Media Propelled courses and course enrichment activities that introduce students to Computational Thinking. iMPaCT integrates exploration of math and programmed computation by engaging students in the design and modification of tiny programs that render raster graphics and simulate familiar kinematics. Through these exercises, students gain experience and confidence with foundational math concepts necessary for success in STEM studies, and an understanding of programmed computation.

In this talk, I describe iMPaCT and present early results from our formal evaluation of semester-length iMPaCT courses indicating improved academic success in concurrently and subsequently attended math courses. They also indicate changes to the nature of student engagement with problem solving using mathematics. I also describe iMPaCT-STEM, a nascent effort of computer science,mathematics, and electrical engineerng faculty to distill iMPaCT's pedagogy into sequences of short learning activities designed to teach and reinforce a variety of mathematical and kinematic concepts that can be directly integrated into math and science courses.

Friday July 8, 2011 -- 12pm in CS221

Gozde Ulutagay, Department of Computer Engineering, Izmir University, Turkey

Title: An Overview of Fuzzy and Crisp Clustering Algorithms

Abstract: Data mining is a modern and crucial technology which leads to effective results by means of bringing mathematical methods and computerized data analysis together. In scientific journals, data mining, together with nanotechnology, biotechnology, and some other fields of technology, is among the most efficient 10 technologies that changes the world. Among the vital tools of data mining, perhaps, clustering, the process of grouping a set of objects into classes of similar objects is the most important one. It has its roots in many areas, including statistics, data mining, biology, machine learning, etc. Cluster analysis is an important human activity. Early in childhood, one learns how to distinguish between cats and dogs, or between animals and plants, by continuously improving subconscious clustering schemes. By clustering, one can identify dense and sparse regions, and therefore, discover overall distribution patterns and interesting correlations among data attributes. Clustering problems require significant infrastructure both mathematically and algorithmically. The aim of this presentation is to address the mathematical and algorithmic aspects of crisp and fuzzy clustering techniques and to illustrate computer applications.

Bio: Gozde Ulutagay received her B.Sc. and M.Sc. degrees from Department of Statistics in Ege University, Izmir, Turkey in 2001 and 2004, respectively. She received her PhD degree from Department of Statistics, Dokuz Eylul University in 2009. Her main research area is fuzzy cluster analysis. She has a plenty of co-authored articles and presentations in the area. Her application of fuzzy neighborhood based clustering in the analysis of Bispectral Index of EEG data was the winning presentation in 2009. She is also interested data mining, optimization, and multivariate statistics. Dr. Ulutagay is currently working as an Assistant Professor in the Department of Industrial Engineering, Izmir University, Turkey. She is also the Vice Dean of The Faculty of Engineering.

Friday May 27, 2011 -- 12pm in CS308

Thomas Chou, UCLA

Title: Stochastic models of viral entry kinetics and inverse problems

Abstract: We develop and study a kinetic model to study the physics of viral infection via the two main entry pathways into cells: fusion and endocytosis. Analysis of the model allows us to derive a ``phase diagram" that yields qualitative predictions of the biophysical conditions under which each entry pathway is preferred. We will also discuss several extensions, all motivated by examining recent experimental protocols introduced to determine the receptor-coreceptor usage in HIV-1 infectivity. In the second part of the talk, I will discuss stochastic inverse problems and present some results on the reconstruction of drift functions in a Brownian motion, as well as on the branching number distribution in a Bellman-Harris branching process. Conditions and additional measurements that render the reconstruction better conditioned, when considering perfect data, will be outlined.

Bio: TBA

Friday May 6, 2011 -- 12pm in CS308

Christelle Jacob, IRIT, France

Title: Uncertainty handling in quantitative BDD-based fault-tree analysis by interval computation

Abstract: In fault-tree analysis probabilities of failure of components are often assumed to be precise. However this assumption is seldom verifi ed in practice. There is a large literature on the computation of the probability of the top (dreadful) event of the fault-tree, based on the representation of logical formulas in the form of a binary decision diagram (BDD). When probabilities of atomic propositions are ill-known and modelled by intervals, BDD-based algorithms no longer apply to the computation of the top probability interval. This paper investigates this question, and proposes an approach based on interval methods, relying on the analysis of the structure of the Boolean formula representing the fault-tree. The considered application deals with the reliability of aircraft operations.
Here are the slides of Christelle Jacob's presentation.

Bio: Christelle Jacob is a PhD student a the "Institut Supérieur de l'Aéronautique et de l'Espace" in Toulouse, France. She works under the supervision of Janette Cardoso and Didier Dubois on a project, @MOST, with Airbus and three other french labs : ONERA, LAAS and IRIT. She is studying the uncertainty management of models for preventive maintenance of aircrafts.

Friday April 29, 2011 -- 12pm at UGLC room 110

Suvrajeet Sen, Ohio State University

Title:A Prognosis for Stochastic Combinatorial Optimization

Abstract: By and large, the optimization literature focuses on deterministic problems, and the corresponding algorithms typically seek solutions that are very fine tuned to the specific data for the model. However, there are many real-world problems in which data is not quite certain, although one might have postulate a probabilistic description for the data. These lead to stochastic optimization models, and when the underlying choices involve combinatorial choices, these problems are stochastic optimization problems. They arise in numerous applications, especially, in situations where data becomes available over time. We will describe some applications from homeland security, and defense. Thereafter, we will discuss some optimal seeking methods based on decomposition algorithms. We will also provide computational evidence that in the face of uncertainty, such algorithms provide a much more realistic computational avenue than traditional deterministic approaches.

Bio: Suvrajeet Sen is Professor of Industrial and Systems Engineering and Director of the Center for Energy, Sustainability, and the Environment. Prior to joining OSU, he served on the faculty at the University of Arizona, and he also served as a program director at NSF where he was responsible for the Operations Research, and the Service Enterprise Engineering programs. Professor Sen is a Fellow of INFORMS. He has served on the editorial board of several journals, including Operations Research as Area Editor for Optimization, and as Associate Editor in INFORMS Journal on Computing, and Journal of Telecommunications Systems. Professor Sen is the past-Chair of the INFORMS Telecommunications Section and founded the INFORMS Optimization Section.

Friday April 8, 2011 -- 12pm in CS221

Michael McGarry, ECE Department, the University of Texas at El Paso

Title: Feed Forward Bandwidth Indication (FFBI): An Unconventional Approach to Bandwidth Forecasting

Abstract: Bandwidth forecasts can empower network protocols with a new intelligence that can bring packet switched networks to new levels of efficiency. With 90% of network traffic projected to consist of video information, our focus is on video bandwidth forecasting. We exploit the fact that for pre-recorded video, the size of every video frame is known prior to the video being delivered through the network. We propose Feed Forward Bandwidth Indication (FFBI) which feeds video frame sizes forward in a sequence of video frames. We extend FFBI to live video by introducing a delay at the source equivalent to the forecast window. We compare FFBI to the most accurate forecast methods found in the literature and use FFBI to improve network performance measures in Ethernet Passive Optical Networks (EPONs). The use of FFBI can provide a 50% reduction in queueing delay compared to the use of no forecasting and a 35% reduction in queueing delay compared to other forecasting methods.

Bio: Michael P. McGarry is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Texas at El Paso. He received his B.S. in Computer Engineering from Polytechnic University, Brooklyn, NY in 1997. He received his M.S. and Ph.D. in Electrical Engineering from Arizona State University, Tempe, in 2004 and 2007, respectively. He is the recipient of the 2009 IEEE Communications Society Best Tutorial Paper award. His research interests include bandwidth forecasting, congestion control, and the optimization of MAC protocols.

Friday March 25, 2011 -- 12pm in CS221

Jeffrey Rickman, Lehigh University
Joint Seminar with UTEP's Department of Metallurgical and Materials Engineering

Title: Some Applications of Computer Simulation in Physics and Engineering

Abstract: I will discuss two illustrative examples in which computer simulation provides insight into problems in physics and engineering. First, I will describe how the generation of weighted Voronoi diagrams and their analysis employing the methods of stochastic geometry are used to model microstructural evolution associated with a phase transformation. In conjunction with these simulations, the (reverse) Monte Carlo method is employed to deduce the nucleation and growth conditions that lead to a particular microstructure. Second, I will describe the modeling of plasticity in metals using level-set and other methods. Plastic deformation involves the motion of line defects (i.e., dislocations), and these methods permit the description of complex topological changes and defect interactions. From the results of simulation, I will then identify operative strengthening mechanisms.

Bio: Dr. Rickman is a Professor of Materials Science and Engineering and Professor of Physics. He joined the Lehigh University faculty in 1993 after postdoctoral appointments at the University of Michigan and Argonne National Laboratory. He did his graduate work in physics at Carnegie Mellon University and his undergraduate work in physics and mathematics at Miami University. His many research interests include the development of computer simulation methodologies for describing fluids and solids, determination of the elastic properties of defects and the kinetics of phase transformations. He has received several honors including the NSF Young Investigator Award, the Chambers Junior Professorship and the Culler Prize (Miami University). He is also a member of several honor societies including Phi Beta Kappa, Phi Kappa Phi and Sigma Xi and several professional organizations including the Materials Research Society, the TMS (Chemistry and Physics of Materials Committee) and the American Ceramics Society.

Thursday March 24, 2011 -- 2pm in CS308

Jan Sliwka, ENSTA-Bretagne, France

Title: Robust localization of underwater robots and other issues in the design of such devices

Abstract: Each year our engineering school ENSTA Bretagne participates in a European competition of autonomous underwater robots called SAUC'E (Student Autonomous Underwater Challenge Europe). The purpose of this competition is to build an intelligent robot capable of performing many missions, such as passing through an underwater gate, following a pipeline, detecting and manipulating underwater objects..., without the help of a human operator. In my talk, I will present our submarine robot "Sauc'isse". I will talk about its mechanical design, electronic design, and software design as well as the different algorithms needed to perform well during the competition. I will explain in more details the localization algorithm since it is part of my PhD work.

Friday February 25, 2011 -- 12pm in CS221

Vladik Kreinovich and Luc Longpre, the University of Texas at El Paso

Title: A Firm Foundation for Private Data Analysis

Abstract: Everyone understands that there is a need to preserve privacy, but it is not easy to precisely define what it means -- and, in situations when some privacy was lost, how to gauge the amount of lost privacy. For example, a person may not want to disclose his or her salary; in this case, disclosing one of the decimal digits of the annual salary amount is a violation of privacy. Disclosing the first digit of the salary does provide a lot of information, but disclosing the last digit seems almost harmless. In this talk, we present a recent survey paper on privacy definitions and measures written by Cynthia Dwork from Microsoft Research. If time allows, we will also explain limitations of the existing approaches -- and our ideas on how these limitations can be overcome.

Friday February 18, 2011 -- 12pm in CS221

Larry Hall, University of South Florida

Title: Finding the right genes for Disease and Prognosis Prediction

Abstract: It is possible to get gene expression data relatively inexpensively from micro-arrays. So, this leads to the possibility that the genes can tell us who will get or has a disease. Perhaps one can find the stage of the disease to enable effective treatments. However, we are currently at the stage where there are many challenges to evaluating the possibilities for genes to be used in diagnosis and treatment. There are typically many more genes that might be involved than samples for any given disease. Which genes are important and how stable are the choices an algorithm provides? We do not know the time of the true onset of a disease, just sometimes when symptoms started and sometimes when diagnosis was done. Some of the promising work on diagnosis or prognosis has suffered from data or scientific errors. This talk will discuss the problems and pitfalls of using genes to predict disease presence or prognosis. It will also discuss some promising ways to choose the genes that may be predictive for a particular disease, with a focus on cancer, and point out some open questions.

Bio: My research interests lie in distributed machine learning, extreme data mining, bioinformatics, pattern recognition and integrating AI into image processing. The exploitation of imprecision with the use of fuzzy logic in pattern recognition, AI and learning is a research theme. He has authored or co-authored over 65 publications in journals, as well as many conference papers and book chapters. Some recent publications appear in the IEEE Transactions on Pattern Analysis and Machine Intelligence, Neural Computation, Information Fusion, Journal of Machine Learning research, IEEE Transactions on Systems, Man, and Cybernetics, Pattern Recognition, the International Conference on Pattern Recognition, the Multiple Classifier Systems Workshop, and the FUZZ-IEEE conference. I co-edited the 1994 joint North American Fuzzy Information Processing Society (NAFIPS), IFIS and NASA conference proceedings and the 1998 proceedings. I am a fellow of the IEEE. I'm a past president of NAFIPS. Also, associate editor for the IEEE Transactions on Fuzzy Systems, International Journal of Approximate Reasoning, International Journal of Intelligent Data Analysis, and The Handbook of Fuzzy Logic. I'm a Fellow of the IEEE. I'm the former Editor-in-Chief for the IEEE Transactions on Systems, Man and Cybernetics, Part B. I am the Jr. Past President of the IEEE Systems, Man and Cybernetics Society.

Thursday December 9, 2010 -- 11am in CS221

Manish Jain, University of Southern California

Title:Game Theory for Security: Algorithms and Applications

Abstract: Predictable allocations of limited security resources such as police officers, canine units, or checkpoints are vulnerable to exploitation by attackers. Game theory provides a principled way to find optimal randomized security policies that thwarts this predictability. In this talk, I will describe game-theoretic models and algorithms that we have developed for security domains. I will also briefly describe the application of some of these algorithms by the Los Angeles Worlds Airport (LAWA) police and the Federal Air Marshals Service (FAMS).

Bio: Manish Jain is currently a PhD candidate at the University of Southern California. He is a part of the Teamcore Research group, led by Prof. Milind Tambe. His work is on the applications of game-theoretic and large-scale optimization techniques, including the scheduling of flights/air marshals for the Federal Air Marshals Service (FAMS) and the scheduling of checkpoints for the Los Angeles International Airport (LAX) police. He has co-authored papers on the subject of security games that have been presented in major artificial intelligence and operations research conferences. His work published in Interfaces was a finalist for the EURO excellence in Practice award. He has also received a Letter of Commendation from the city of Los Angeles for his contributions to the development of the security assistant deployed at LAX.

Thursday November 11, 2010 -- 3pm in Bell Hall 143

Irina Perfilieva and Vilem Novak, University of Ostrava, Czech Republic

Title: Discrete Fuzzy Transforms and their Applications in Data Processing

Abstract: The theory of fuzzy (F-)transforms can be seen as a bridge between fuzzy modeling and the theory of linear (in particular, integral) transforms. In this talk, we will show backgrounds of the ordinary F-transform and a higher order F-transform. We will explain different ways of achieving a desired quality of approxima- tion by the, so called inverse F-transform. Various applications demonstrate the universality of this special technique of fuzzy modeling. We will mention two of them: image processing, and analysis and forecast- ing of time series. The talk will be self-contained, illustrated by pictures and provided with all necessary explanations.

Friday November 12, 2010 -- noon to 1pm, in CS308

Huiping Cao, New Mexico State University

Title: Feedback-driven Result Ranking and Query Refinement for Exploring Semi-structured Data Collections

Abstract: Feedback process has been used extensively in document-centric applications, such as text retrieval and multimedia retrieval. Recently, there have been efforts to apply feedback to semi-structured XML document collections as well.
In this work, I note that feedback can also be an effective tool for exploring (through result ranking and query refinement) large semi-structured data collections. In particular, in large scale data sharing and curation environments, where the user may not know the structure of the data, queries may initially be overly vague.
Given a path query and a set of results identified by the system to this query over the data, I consider two types of feedback: Soft feedback captures the user’s preference for some features over the others. Hard feedback, on the other hand, expresses users’ assertions regarding whether certain features should be further enforced or, in contrast, are to be avoided. Both soft and hard feedback can be “positive” or “negative”. For soft feedback, I develop a probabilistic feature significance measure and describe how to use this for ranking results in the presence of dependencies between the path features. To deal with the hard feedback efficiently (i.e., fast enough for interactive exploration), I present finite automata based query refinement solutions. In particular, I present a novel LazyDFA+ algorithm for managing hard feedback. I also describe optimizations that leverage the inherently iterative nature of the feedback process. I bring together these techniques in AXP, a system for adaptive and exploratory path retrieval. The experimental results show the effectiveness of the proposed techniques.

Bio: Dr. Huiping Cao received a Ph.D. in Computer Science from The University of Hong Kong in 2007, a Master's in Computer Science and a Bachelor's in Management Information Systems from Renmin University of China in 2002 and 1999 respectively. She is an assistant professor in Computer Science at New Mexico State University (NMSU). Before joining NMSU, she worked as a Research Fellow at University of California Santa Barbara and as a Postdoctoral Research Associate at Arizona State University. Huiping's research interests are broadly in the area of data management (e.g., data discovery, indexing and integration) and mining of non-traditional data (e.g., spatial, spatiotemporal and scientific data). She has published articles on data management and data mining in highly competitive venues.

Friday October 29, 2010 -- noon to 1pm, in CS308

Chris Kiekintveld, CS Department,UTEP

Title: Overview of research interests

Abstract: I will present a broad overview of my research and interests in the area of artificial intelligence, multi-agent systems, and strategic reasoning. I am interested in multi-agent decision problems, which include both traditional games such as chess or poker, as well as a wide variety of important real-world problems that can be modeled as games. For example, I am currently working on several project that use game theory to help make complex resource allocation decisions in security domains. One of these projects developed a software tools that is curently being used by the Federal Air Marshals Service (FAMS) to help create unpredictable, risk-based flight schedules for the air marshals. Border security is a new and exciting area where I am looking to apply similar methods. Finally, I will briefly describe some of my other work on developing intelligent trading agents, and on algorithms for distributed optimization.

Cuauhtemoc Munoz, CS Department,UTEP

Title:Automated Testing of LTL Formulas with Execution Traces using Prospec

Abstract:Prospec is a tool that allows software engineers to create software specifications by combining patterns, scopes, and composite propositions (CPs). Patterns, scopes, and CPs are based on Linear Temporal Logic (LTL) formulas. Salamah designed a general algorithm for generating LTL formulas from patterns, scopes, and CPs, and Vela implemented the algorithm. The LTL generator is capable of generating LTL formulas for every pattern, scope, and CP combination. While there are at least 30,000 combinations of patterns, scopes and CPs, only 164 LTL formulas were manually tested in the implementation.
PROTEF is a framework for testing LTL formulas using execution traces. An execution trace represents a sequence of states, where each state is defined by a set of true atomic propositions. PROTEF takes an execution trace and an LTL formula and generates a model that can be checked using the NuSMV model checker. The model checker determines whether the formula is satisfied by the model of the execution trace. This talk describes the LTL Verifier, which automatically generates test cases to verify the LTL Generator. The approach to test case generation and the test oracle are novel. Test cases are composed of LTL formulas, execution traces, and expected results. Execution traces and expected results are automatically generated by the LTL Verifier. Test oracles predict the correct results. Test cases are generated by considering equivalence classes and boundary values and applying specific rules to patterns, scopes, and CPs. Test cases generated by the LTL Verifier are executed using PROTEF. Approximately 2% of the 3,836,960 test cases failed. Two types of errors were found in the LTL formulas by using the LTL verifier. The first type of error was missing parentheses in the LTL formulas. The second type of error was a wrong implementation of the LTL algorithm for a specific pattern, scope, and CP combination.

Luc Longpre, CS Department, UTEP

Title: Security and Privacy

Abstract: An important part of security is protection of confidentiality. We explore different aspects of privacy, including definitions, privacy in statistical databases, escrowed privacy.

Friday October 15, 2010 -- noon to 1pm, in CS221

Guoqiang Hu, CS Department, UTEP

Title: A Full Life-Cycle Methodology for Structured Use-Centered Quantitative Usability Requirements Specification and Usability Evaluation of Websites.

Abstract: World Wide Web has gained its dominant status in the cyber information and services delivery world in recent years. But how to specify website usability requirements and how to evaluate and improve website usability according to its usability requirements specification are still big issues to all the stakeholders. To help solve this problem, we propose a website usability requirements specification and usability evaluation methodology that features a structured use-centered quantitative full life-cycle method. A validation experiment has been designed and conducted to prove the validity of the proposed methodology, QUEST (Quantitative Usability Equations SeT). Its principle is to prove that QUEST has stronger website usability evaluation capability than the most typical existing usability evaluation methods. Apparently, if QUEST's website usability evaluation capability is established, then its usability metrics can be used to quantitatively specify upfront user usability requirements for websites. In the validation experiment, 7 usability experts and 20 student subjects were recruited to perform 4 tasks on 2 open source calendar websites, WebCalendar 1.0.5 and VCalendar 1.5.3.1; 4 sets of usability data had been collected, which were corresponding to the following 4 usability evaluation methods respectively: expert usability review, traditional user usability testing, SUS (System Usability Scale), and QUEST. According to the experiment results: both the expert usability review and the traditional user usability testing were inconclusive on which of the 2 target websites had better usability; although SUS rated the overall usability of WebCalendar 1.0.5 at 66.00 and VCalendar 1.5.3.1 at 61.75, it was subjective and vague on usability problems; in contrast, QUEST not only rated the overall usability of WebCalendar 1.0.5 at 56.59 and VCalendar 1.5.3.1 at 35.97, but also revealed where the usability problems were and how severe each usability problem was in a quantitative manner. In conclusion, it clearly can be stated that QUEST has stronger website usability evaluation capability than all other 3 most typical existing usability evaluation methods. So, the proposed methodology has been validated by the experiment results.