ºÚÁϳԹÏÍø

2015-16 Monthly Meetings

1. August 23: 3:00-5:00 p.m. (Stright 240)
2. September 29: 4:30-6:00 p.m. (Stright 240)
   
3. October 28: 6:30-8:00 p.m.(Stright 240)
4. December 4: 3:30-5:00 p.m.(Stright 240)
5. January 24, 3:00-5:00pm(Stright 240)
6. Feb. 29, 6:30-8:00pm(Stright 240)
7. Mar. 31, 3:30-5:00pm(Stright 229)
8. Apr. 29, 3:30-5:00pm(Stright 240)

Workshops: (Open to Public)

1. Workshop: Introduction to R -- Part I
   3:30-5:00 p.m., September 22, in Stright 220
   Presenter: Dr. Russ Stocker
   
2. Workshop: Introduction to R -- Part II
   3:30-5:00 p.m., September 24, in Stright 220
   Presenter: Dr. Russ Stocker
   Abstract: R is a free statistical software package that is used extensively in both academia and industry. It is an open source platform with currently more than 7,000 contributed packages that include the latest statistical tools for data analysis. In this two-day workshop, we introduce the major components of R. These include importing and exporting data, data manipulation, descriptive statistics, graphics, inferential statistics, and statistical modeling. Participants in the workshop will be provided handouts that contain R examples and they will use R to complete a variety of exercises designed to help them learn the package.
3. LINGO Workshop
   3:30-5:00pm, Feb. 3, in Stright 220
   Presenter: Dr. John Chrispell
4. SAGE Workshop
   3:30-5:00pm, Mar. 2, 3:30-5:00pm, in Stright 220
   Presenter: Dr. John Chrispell
5. 3D printer workshop
   3:00-5:00pm Mar. 30, in Stright 220
   Presenter: Dr. Brian Sharp and Dr. Ed Donley

Invited Speakers: (Open to Public)

1. October 12, 2015
   1:25-2:15 p.m. Pratt Auditorium
   Presenter:Jeremy Yagle, Data Analyst, NASA Langley Research Center
   Title:Temporal Pattern Mining for the Prediction of Flutter from Aeroelasticity Data
   Abstract:Aeroelastic Flutter is a highly non-linear phenomenon caused by a combination of aerodynamics, inertial forces, and the elastic properties of the wing of an aircraft. Researchers at NASA are working to advance fundamental knowledge of aeroelastic phenomena through the development and application of prediction methods to aerospace vehicles. In support of their work, we are investigating a data-driven approach to understanding flutter by applying temporal pattern mining algorithms to the experimental data collected during wind tunnel testing. When testing an aircraft model in a wind tunnel, scientists and engineers currently rely on expert observation, monitoring of instrumentation, and the use of Fourier Transforms to convert data from the time domain to the frequency domain. This approach may not always be effective in identifying subtle dynamics that are precursors to flutter.
   In this talk, I will discuss the use of pattern mining algorithms to identify repetitive subsequences - or "signatures" - in time series data collected from accelerometers. The family of algorithms being investigated has the capability to mine unprocessed, time-domain data in a fraction of the time required by other methods. Once identified, these signatures can then be mapped to events that occurred during wind tunnel testing. Detection and identification of significant signatures could lead to new insights about precursors to flutter and other aeroelastic phenomena. An overview of the algorithms and test methodology will be presented, along with a discussion of the potential challenges and benefits to this important area of aerospace research.
   About the presenter:
   

   At NASA Langley Research Center, I am involved in the Comprehensive Digital Transformation, which is an initiative focused on developing and implementing innovative data analytics and machine intelligence solutions for complex problems in Langley's aerospace domain. Our work is centered on two key areas:

   Data Intensive Scientific Discovery:Automated mining of images, experimental data, and computational data that will help NASA scientists derive new insights and make new discoveries. The tools we develop in this area incorporate the underlying physics into the machine learning algorithms.

   Deep Content Analytics:Knowledge mining of scientific literature, web content, and multimedia sources that will allow NASA researchers to quickly analyze vast collections of information in order to answer specific questions. The tools in this area rely on natural language processing and IBM Watson technologies.

   The overall goal of our work is to provide NASA scientists with new tools that will lead to greater scientific discoveries and system design optimizations. While I am involved with several different project across the two key areas, the primary focus of my research is on the detection and prediction of aeroelastic flutter from wind tunnel test data.

   b. Career Panel: 3:30pm-5:00pm in Stright Hall Room 226/229
   
   Panel:
   Nathan Adelgren, Ph.D. candidate, Clemson University
   Ryan Grove, Ph.D. Program at Clemson University & Technical Staff at The
   Aerospace Corporation
   Matthew Sulkosky, Software Developer, Technology Management Associates
   Jeremy Yagle, Data Analyst, NASA Langley Research Center
   
   Biography:
   Nathan Adelgren:I'm originally from Jamestown, NY. I attended IUP from 2009 to 2011 and earned a Master's degree in Applied Math. I am now beginning my fifth, and hopefully final, year working on a PhD at Clemson University in Clemson, SC. While at IUP I particularly enjoyed the courses I took in Operations Research with Dr. Kuo, and Computational Math with Drs. Kuo, Adkins and Donley. I chose to attend Clemson because it was one of the few Math programs that I found which offered classes in these areas in additional to the more common pure math areas, algebra and analysis.
   I have completed all of my degree requirements at Clemson except the completion of my dissertation, which will cover the work I've done in two distinct areas of Operations Research: Multiobjective mixed-integer programming and multiparametric linear complementarity problems. I will now briefly explain my research, one topic at a time.
   Multi-objective mixed-integer programming:Multiobjective optimization is a relatively new field in which optimization problems are studied that have more than one objective function. For example, you could consider maximizing profit and customer satisfaction simultaneously. As these objectives are often conflicting, one can no longer expect to find a single "optimal solution," but rather must look for a set of "Pareto-optimal solutions," which can be thought of as the solutions that offer an acceptable compromise between the objectives. Researchers have developed many tools for solving these problems when all the variables in a model are continuous, but relatively little work has been done for problems containing integer variables. I am using the C programming language alongside IBM's optimization package, CPLEX, in order to develop a solver for problems containing both integer and continuous variables and two objectives.
   Multi-parametric linear complementarity problems:Multiparametric optimization is similar to the stochastic optimization that you've likely seen in your OR courses, but in this case some of a problem's defining data is simply assumed to be unknown rather than assumed to follow some probability distribution. The goal is to find the optimal solution as a function of the parameters. Linear programs (LPs) and quadratic programs (QPs) can both be reformulated into a problem known as the linear complementarity problem (LCP) and thus, studying multiparametric LCP is a way of studying multiparametric LPs and QPs. Much work has been done in the past to determine methods for solving multiparamtric LPs and QPs with parameters in the coefficients of the linear terms of the objective function and/or the right hand sides of the constraints, but little work has focused on when parameters occur in arbitrary locations.My research consists of developing the theory necessary for solving these problems with parameters in arbitrary locations.
   Ryan Groveis currently a part-time Member of the Technical Staff (MTS) atThe Aerospace Corporation.He is also a Graduate Teacher of Record, teachingMATH1060atClemson University. This is in addition to being a 3rd year PhD student and doing research under his advisor,Timo Heister, where hesolves advection-diffusion equations that appear in the finite element discretization of a mantle convection simulation.In the summer of 2015, he was a Member of the Technical Staff (MTS) Summer PhD atThe Aerospace Corporation, where he was a developer for the Genetic Resources for Innovation and Problem Solving (GRIPS) program, which is a decision-support process that uses evolutionary algorithms, efficient parallel processing on thousands of compute cores, and advanced high-dimensional visualization to solve complex problems.
   Matthew Sulkosky:I grew up very close to the university in the town of Blairsville. After graduating high school, I joined the ºÚÁϳԹÏÍø Army National Guard in July of 2009. I completed initial training and returned to ºÚÁϳԹÏÍø to enroll in college. I started at Penn State University for Electrical Engineering. My first year there I learned that I strongly disliked my choice, and transferred to ºÚÁϳԹÏÍø for my two favorite subjects: Mathematics and Computer Science. Two years and an internship later, I graduated from IUP with two bachelors degree and moved to the Virginia area to work as a software developer. I have worked on a few different projects since moving to the area, mostly in the field of web development and a few projects in data analytic. This has been over the course of working at three different companies: Booz Allen Hamilton, Information Innovators Inc., Technology Management Associates.

2. Feb.23, 2016
   , , Mathematical Modeling Group, National Institute of Standards and Technology

   1. 11:00-11:50 a.m. Pratt Auditorium

      Title: A Mathematical Look at Paint, Hollywood, and Networks

      Abstract:

      This talk will present several examples of the kinds of problems a mathematician can encounter at NIST. They illustrate the breadth of applications mathematical ideas.

   2. 3:30-4:30 p.m. STRGT 226/22

      Title: An Algorithm for Identifying Optimal Spreaders in a Random Walk Model of Network Communication

      Abstract:

      In a model of network communication based on a random walk in an undirected graph, what subset of nodes (of some fixed size), enable the fastest spread of information? The dynamics of spread is described by a process dual to the movement from informed to uninformed nodes. In this setting, an optimal set A minimizes the sum of the expected first hitting times (F(A)), of random walks that start at nodes outside the set. Identifying such a set is a problem in combinatorial optimization that is probably NP hard. Fortunately, F has been shown to be a supermodular and non-increasing set function.

      In this talk, the problem is reformulated so that the search for solutions to optimization problem is restricted to a class of optimal and "near" optimal subsets of the graph. We will discuss our approach to the approximation and solution of this problem based on properties of the underlying graph.