SPEAKER: Dr. Zhaohui Steve Qin

ABSTRACT:

Understanding how chromosomes fold provides insights into transcription regulation hence functional state of the cell. Recently, chromosomal conformation capture (3C)-related technologies have been developed to provide a genome-wide view of chromatin organization. Despite great technologies, multiple layers of noise and uncertainties stem from the sophisticated experiments, coupled with various sequencing-related artifacts, making the analysis of such data extremely challenging. Here using Hi-C as an example, we review the critical issues of analyzing this latest type of genomics data, including normalization, modeling and inference. We describe a novel Bayesian probabilistic approach, denoted Bayesian 3D constructor for Hi-C data(BACH), to infer chromosome three-dimensional (3D) structures from Hi-C data. We also discuss the observations we made when applying BACH to real Hi-C datasets. This is a collaboration with Ming Hu, Ke Deng, Jesse Dixon, Siddarth Selvaraj, Jennifer Fang, Bing Ren and Jun Liu.

Contact: Zhaohui Steve Qin <zhaohui.qin@emory.edu>

BIO:  Dr. Qin is currently an Associate Professor in the Department of Biostatistics and Bioinformatics at Rollins School of Public Health, Emory University. He is also a faculty member at the Department of Biomedical Informatics, Emory University School of Medicine and Biostatistics and Bioinformatics Shared Resource, Winship Cancer Institute. Dr. Qin received his B.S. degree in Probability and Statistics from Peking University in 1994 and Ph.D. degree in Statistics from University of Michigan in 2000. He was a postdoctoral fellow in Dr. Jun Liu's group in Department of Statistics at Harvard University from 2000 to 2003. He joined the Department of Biostatistics at University of Michigan in 2003. In 2010, he moved to his current position in Emory University. Dr. Qin has more than ten years of experience in statistical modeling and statistical computing with applications in statistical genetics and genomics. Recently, his research is focused on developing Bayesian model-based methods to analyze data generated from applications of next generation sequencing technologies such as ChIP-seq, RNA-seq and resequencing. Dr. Qin also actively collaborates with biomedical scientists and clinicians on various projects that utilizing next generation sequencing technologies to study cancer genomics. Dr. Qin has published more than 50 peer-reviewed research papers covering statistics, bioinformatics, statistical genetics and computational biology. He has severed multiple times as ad hoc member on various NIH study sections and has supervised more than 10 graduate students and postdoctoral fellows.

SPEAKER: Prof. Peihua Qiu

ABSTRACT:

In our daily life, we often need to identify individuals whose longitudinal behavior is different from the behavior of those well-functioning individuals, so that some unpleasant consequences can be avoided. In many such applications, observations of a given individual are obtained sequentially, and it is desirable to have a screening system to give a signal of irregular behavior as soon as possible after that individual's longitudinal behavior starts to deviate from the regular behavior, so that some adjustments or interventions can be made in a timely manner. We propose a statistical process screening system for that purpose, using statistical process control and longitudinal data analysis techniques. Our proposed method is demonstrated using a real-data example about the SHARe Framingham Heart Study of the National Heart, Lung and Blood Institute.

Contact: Peihua Qiu <qiuxx008@umn.edu>

BIO:

Dr. Qiu got his Ph.D. in statistics from the Statistics Department at University of Wisconsin at Madison in 1996. He worked as a senior research consulting statistician of the Biostatistics Center at Ohio State University during 1996-1998. Then, he worked as an assistant professor (1998-2002), an associate professor (2002-2007), and a full professor (2007-present) of the School of Statistics at University of Minnesota. He is an elected fellow of the American Statistical Association, an elected fellow of the Institute of Mathematical Statistics, an elected member of the International Statistical Institute, a senior member of the American Society for Quality, and a lifetime member of the International Chinese Statistical Association. He served as associate editor for Journal of the American Statistical Association (2006-2012), Biometrics (2011-2012),Technometrics (2007-2012), Statistical Papers (2011-2012), and guest co-editor for Multimedia Tools and Applications. He will be the Editor-Elect (2013) and Editor (2014-2016) of Technometrics.

SPEAKER:  Mark Lewis

ABSTRACT:

In this talk, I will present several classical decision-making paradigms revised for today's environment. In particular, in the service economy models of call centers and health care systems include abandonments. Assuming customer patience is exponential, modeling basic systems with continuous-time Markov processes remains valid, but analyzing them using uniformization is not. This means that discrete-time Markov decision process theory (and the accompanying successive approximations) cannot be used. Similarly, in health care systems arrivals to emergency departments are rarely stationary. We mention how to extend MDP theory to more general state spaces. Finally, can the two challenges be met together? Open problems are discussed.

BIO:

Professor Lewis received his Ph.D. in Industrial and Systems Engineering from the Georgia Institute of Technology in 1998. After receiving his doctorate, Lewis spent a year at the University of British Columbia as a postdoctoral fellow in the Center for Operations Excellence. He joined Cornell's School of Operations Research and Information Engineering in 2005 after teaching Industrial and Operations Engineering at the University of Michigan.

Broadly speaking Professor Lewis's research interests are on the dynamic control of service systems. Most often he uses the methodology of stochastic dynamic programming or Markov decision processes to analyze these problems. Along the way he has done fundamental research on the theory of MDPs. In the area of average cost MDPs on general state and action spaces he has studied convergence of discounted cost optimal values and policies and on a refinement of average cost theory called bias optimality he has studied implicit discounting. In terms of applications, Professor Lewis has considered routing in transportation systems, control of inventory systems and allocation of inter-switch handoffs in wireless communications. Despite his versatility in the analysis of such systems his passion is for resource allocation in controlled queueing networks. In doing so, he has considered non-stationary networks, networks with limited capacity and those with varying service capabilities.

SPEAKER: Professor T.N. Sriram

ABSTRACT:

In this talk, we will introduce a family of multivariate association measures based on density power divergences that help recover both linear and nonlinear relationships between multiple sets of random vectors. This approach not only characterizes independence, but also provides a smooth bridge between well-known distances that are inherently robust against outliers. Algorithmic approaches are developed for dimension reduction and selection of the optimal robust association index. Extensive simulation studies are performed to assess the robustness of these association measures under different types and proportions of contamination. We illustrate the usefulness of our methods in applications. Some theoretical properties, including the consistency of the estimated coefficient vectors, are investigated and computationally efficient algorithms for our nonparametric methods are provided.

SPEAKER: Pierre Le Bodic

ABSTRACT:

In classical cut problems, the input consists of a graph and some set S to cut. Partial cut problems are a generalization of cut problems, in which, given an additional integer k, only a subset of S of cardinality at least k must be cut. The archetypal example is the partial multicut problem, in which S is a set of pairs of vertices. The focus of this article is on partial cut problems for which the classical version is easily solvable on some class of graphs. A variant of multiterminal cut is proven to become \NPhard{} when its partial version is considered. The major part of this talk is dedicated to designing a unified dynamic programming algorithm for partial cut problems. Using this result, many partial cut problems are proven to be FPT w.r.t. some parameters including the treewidth of the input graph, or, for the generalized version, pseudopolynomial if these parameters are fixed. A partial cover problem, namely partial dominating set, is also solved using this unified algorithm. Using these algorithms, FPTASs are then designed for generalized partial versions of multicut and multiterminal cut on bounded treewidth graphs. A (2+\epsilon)-approximation is also provided for the generalized partial multiterminal cut problem on general graphs, and adapted, with different ratios, to other variants.

SPEAKER: Dr. Kurt Rohloff

ABSTRACT:

In this talk we discuss recent approaches to modeling and detecting epidemics of Internet worms. We particularly focus on Random Constant Scanning (RCS) worms that use zero-day exploits for which malware signatures are unknowable. Although these worms propagate by randomly scanning network addresses for hosts that are susceptible to infection, traditional RCS worm models are fundamentally deterministic. We introduce a recent modeling approach based on density-dependent Markov jump process to model worm epidemic propagation across the Internet. This model relies on a computationally simple hybrid deterministic/stochastic point-process model for locally observed scanning behavior. We use this behavior as the basis of a worm epidemic detection strategy. We discuss the benefits and drawbacks of a mathematically optimal hypothesis-testing detection approach under idealized conditions and possible other approaches to model, detect and mitigate this class of malware.

BIO: Dr. Kurt Rohloff is a senior scientist in the Distributed Systems research group at BBN Technologies. Dr. Rohloff’s areas of research expertise include supervisory control, stochastic modeling and large-scale distributed computing. Dr. Rohloff is the PI on the DARPA PROCEED program where he leads a team developing a Fully Homomorphic Encryption implementation which will enable secure cloud computing. Dr. Rohloff is the Chief Designer and Lead Architect of the SHARD triple-store, a high-performance, massively scalable graph data storage system. Dr. Rohloff received his MS and PhD degrees from the University of Michigan and was a post-doc at the Coordinated Sciences Lab at UIUC. Dr. Rohloff is very proud of his Bachelor's degree in Electrical Engineering from Georgia Tech (and the ISYE classes on statistical failure modeling.)

SPEAKER: David McDonald

ABSTRACT:

The goal of an on-line quality control procedure is to rapidly detect an out-of-control situation; i.e., to detect a change in the sampling distribution after a change point. We first review a nonparametric Cusum procedure based on the sequential ranks for univariate data.

Many attempts have been made to assign sequential ranks to multivariate data. For instance ranks based on the Tukey depth have been used but the computations are prohibitive in high dimension. Instead we will score observations based on where they fall relative to previous observations.

Each successive observation creates a Voronoi cell indexed by the observation number. Suppose observation $n+1$ falls into the Voronoi cell with index $i$ where $i$ ranges from $1$ to $n$. Then observation $n+1$ has associated score $i/(n+1)$ and becomes the center of a new Voronoi cell with index $n+1$.

Before the change point, the scores are uniformly distributed. After the change point, the scores tend to be large since the observations tend to clump together. We use these scores in a Cusum procedure. We find we can approximately predict the on-target average run length of our procedure, and we get reasonable off-target run lengths for any kind of structural break like a change of mean or a change of dispersion.

BIO: 

David Mcdonald is a Professor in the Department of Mathematics at the University of Ottawa. He received his Ph.D. from U. de Montreal, and a Masters from King's College (London). He has held several visiting positions, including a stay at the Georgia Institute of Technology and Ecole Normale Superieure. He is also a Fellow of the Institute of Mathematical Statistics.

SPEAKER: E. Labner, Asst. Prof., NC State

ABSTRACT:

Penalized maximum likelihood methods that perform automatic variable selection have been developed, studied, and deployed in almost every area of statistical research. A prominent example is the LASSO Tibshirani (1996) with its numerous variants. It is now well-known, however, that these estimators are nonregular and consequently have limiting distributions that can be highly sensitive to small perturbations of the underlying generative model. This is the case even for the fixed “p” framework. Hence, the usual asymptotic methods for inference, like the bootstrap and series approximations, often perform poorly in small samples and require modification. Here, we develop locally asymptotically consistent confidence intervals for regression coefficients when estimation is done using the Adaptive LASSO (Zou, 2006) in the fixed “p” framework. We construct the confidence intervals by sandwiching the nonregular functional of interest between two smooth, data-driven, upper and lower bounds and then approximating the distribution of the bounds using the bootstrap. We leverage the smoothness of the bounds to obtain consistent inference for the nonregular functional under both fixed and local alternatives. The bounds are adaptive to the amount of underlying non- regularity in the sense that they deliver asymptotically exact coverage whenever the underlying generative model is such that the Adaptive LASSO estimators are consis- tent and asymptotically normal, and conservative otherwise. The resultant confidence intervals possess a certain tightness property among all regular bounds. Although we focus on the case of the Adaptive LASSO, our approach generalizes to other penalized methods including the elastic net and SCAD.

SPEAKER: Dr. Joe Qin

ABSTRACT:

Process monitoring provides supervision of process operations so that abnormal operating conditions can be detected, diagnosed, and proper adjustment can be implemented as needed. The ultimate purpose is to reduce process variability under real-world operating conditions with the use of real time data. A focus of this seminar is on the recently rising interest of big data and the use of multivariate statistical methods for efficient data-driven quality-relevant process monitoring. 
 
As an improvement over principal component analysis, a concurrent projection to latent structures is presented for the monitoring of output-relevant faults that affect the quality and input-relevant process faults. The input and output data spaces are concurrently projected to five subspaces, a joint input-output subspace that captures co-variations between input and output, an output-principal subspace, an output-residual subspace, an input principal subspace, and an input-residual subspace. Hotelling’s and residuals-based indices are developed for various fault detection alarms based on these subspaces. The proposed monitoring method offers complete monitoring of faults that happen in the predictable output subspace and the unpredictable output residual subspace, as well as faults that affect the input spaces only.

BIO:  Dr. S. Joe Qin is the Fluor Professor of Process Engineering and Vice Dean at the Viterbi School of Engineering at University of Southern California. He obtained his B.S. and M.S. degrees in Automatic Control from Tsinghua University in Beijing, China, in 1984 and 1987, respectively, and his Ph.D. degree in Chemical Engineering from University of Maryland at College Park in 1992. Dr. Qin’s research interests include statistical process monitoring and fault diagnosis, model predictive control, system identification, run-to-run control, semiconductor process control, and control performance monitoring. He is a Co-Director of the Texas-Wisconsin-California Control Consortium where he has been principal investigator for 17 years. He is a recipient of the National Science Foundation CAREER Award, the 2011 Northrop Grumman Best Teaching award at Viterbi School of Engineering, the DuPont Young Professor Award, Halliburton/Brown & Root Young Faculty Excellence Award, NSF-China Outstanding Young Investigator Award, a Chang Jiang Professor by the Ministry of Education of China from 2007-2010, and an IFAC Best Paper Prize for the model predictive control survey paper published in Control Engineering Practice. He is currently an Associate Editor for Journal of Process Control, IEEE Control Systems Magazine, and IEEE Transactions on Industrial Informatics, and a Member of the Editorial Board for Journal of Chemometrics. He served as an Editor for Control Engineering Practice and an Associate Editor for IEEE Transactions on Control Systems Technology. He is a Fellow of IEEE. He has published over 100 papers in SCI journals, with over 3600 ISI citations and an ISI h-index of 30.

SPEAKER: Major General Ronald Johnson, US Army, (retired)

ABSTRACT:

In this presentation, MG Johnson will discuss three areas of focus - including data gathering, forming the team to optimize performance, and training/development for sustained excellence - using NBA officiating sports analytics as "case study".

MG Ronald Johnson is a candidate for the Director of the Tennenbaum Institute at Georgia Tech and Professor of the Practice in the School of ISyE.

Keynote Address at the External Advisory Board Meeting of the Georgia Tech Manufacturing Institute:
Innovation Imperative and the 21st Century University

Wednesday, October 17, 2012

1:30 -2:30 p.m.

Fuller E. Callaway, Jr. Manufacturing Building Auditorium, 1st Floor

813 Ferst Drive, N.W.

Atlanta, GA.

GT QCF Day

*8:30* Welcome, Dr. Shijie Deng

*8:40 -- 9:25* Dr. Jain, Sandeep, Managing Director, Macro Rates, UBS "A practitioner's approach to model development (with focus on Inflation Modeling)

*9:30 -- 10:15* Dr. Steven Zhu, Director, Credit Risk Management, Morgan Stanley "Financial Crisis and Risk Management -- Lessons and Challenges"

*10:15 -- 10:45* BREAK

*10:45 -- 11:30* Dr. Kevin Kindall, Director, Quant Analysis in Commercial Division, Conoco Phillips "Overcoming Data Shortfalls"

*11:30 -- 12:15* Dr. Philip H. Dybvig, Boatmen's Bancshares Professor of Banking and Finance, Washington University in Saint Louis "The new risk management: the Good, the Bad, and the Ugly"

*12:15 -- 13:40* LUNCH

*13:40 -- 14:25* Ms. Savita Subramanian, Head of US Equity & Quantitative Strategy, Global Macro Research, Bank of America - Merrill Lynch "The Changing Landscape for Quants" *14:30 -- 15:15* Mr. Mark Matthews, Head of Global Quantitative Research, Fixed Income Division, Invesco "Asset Allocation Models and Forecasts"

*15:15 -- 16:00* BREAK

*16:00 -- 16:45* Dr. Premal Shah, VP, Credit Quantitative Analytics, Barclays "Managing Capital Costs for Centrally Cleared Derivative Transactions"

*16:45 -- 17:00* Closing Remarks, Dr. Shijie Deng

Registration is required - please do so at your earliest convenience so we can have an accurate count.

To register and find out more information about the event (speakers' bios, direction to the venue, etc.), go to http://www.qcf.gatech.edu/qcfday/program/index.php

SPEAKER: Dr. Nan Chen, NUS

ABSTRACT:

Queueing Performance Metrics (QPMs) estimate important measures of the system performance, like average cycle time, average waiting length, and throughput rate. These metrics need to be quantitatively evaluated and monitored in real time to continuously improve the system performance. However, QPMs are often highly stochastic, and hence are difficult to monitor using existing methods. In this paper, we propose the cumulative sum (CUSUM) schemes to efficiently monitor the parameters of typical queueing systems. We compared the CUSUM schemes with several alternative methods, and demonstrated that the performance of CUSUM is superior, responding faster to many shift patterns. We also investigate the performance of the CUSUM schemes based on different sampling schemes. We illustrated that more information from complete observations can indeed improve the monitoring performance of CUSUM charts. At last, we used a case study to demonstrate the application of our approach.

Bio:  Dr. Nan Chen is an Assistant Professor in the Department of Industrial and Systems Engineering at National University of Singapore. He obtained his B.S. degree in Automation from Tsinghua University, and M.S. degree in Computer Science, M.S. degree in Statistics, and Ph.D. degree in Industrial Engineering from University of Wisconsin-Madison. His research interests include statistical modeling and surveillance of service systems, condition monitoring and prognostics. He is a member of INFORMS, IIE, and IEEE.

SPEAKER:  Professor Burhan Sandikci

ABSTRACT:

Many hospitals (in particular, academic medical centers) in the United States experience elevated demand and strained inpatient bed supply. This imbalance between the supply and the demand, coupled with intrinsic problems in health care industry, is a major cause of financial struggles as well as a significant challenge to achieve research and education missions in academic medical centers. This talk is motivated by such struggles at the University of Chicago Medical Center (UCMC). To address adverse effects of limited capacity, UCMC started providing care by partitioning the bed capacity into specialized wings. Each wing has a specific designation of the types of patients it can admit, and the number of beds it is allocated. A patient requesting hospital services can be admitted only if a bed is available in the appropriate wing. The fundamental trade-off included in wing formation decisions is between the advantages of pooling in large wings and the advantages of focused care in smaller wings. In this talk, I will present a model to help hospital administrators make wing formation decisions and a novel approach to solve the hard problem of forming wings. I will also share a number of managerial insights through numerical results based on data from UCMC as well as national databases.

BIO:  Burhan Sandikci is an Assistant Professor of Operations Management at the University of Chicago’s Booth School of Business. He received his PhD in industrial engineering in 2008 from the University of Pittsburgh. His research interests span decision-making problems under uncertainty with particular focus on problems in medical decision-making and healthcare operations. His methodological interests include Markov decision processes (MDPs), partially observed MDPs, stochastic games, stochastic programming, and simulation. His research has been published in leading academic journals such as Operations Research, Management Science, and Mathematical Programming. His work has also been recognized at various levels by INFORMS Decision Analysis Society, INFORMS Bonder Scholarship, and IIE Pritsker Doctoral Dissertation Award.

Speakers Include:

Branko Terzic, executive director, Deloitte Center for Energy Solutions
An international consultant to corporations, multilateral lending agencies and governments on energy, infrastructure and network industry issues, Dr. Terzic has extensive experience in the oil and gas industry in valuation, management and regulation having regulated natural gas distribution companies at retail, interstate natural gas and oil pipelines at the FERC and serving as CEO of a natural gas distribution company. Dr. Terzic is a former Commissioner of the U.S. Federal Energy Regulatory Commission (FERC), Commissioner on the State of Wisconsin Public Service Commission (WPSC) and Chairman, CEO and President of the holding company Yankee Energy System Inc. and subsidiary Yankee Gas Services Company (Meriden, CT). Dr. Terzic also serves as Chairman of the United Nations Economic Commission for Europe (UN ECE) and Ad Hoc Group of Experts on Cleaner Electricity Production from Coal and Other Fossil Fuels (Geneva, Switzerland). In 2010 he was appointed to The National Coal Council, the advisory body to the U.S. Secretary of Energy having previously served on the DOE Secretary’s National Petroleum Council.

Valerie Thomas, Anderson Interface Associate Professor of Natural Systems, School of Industrial and Systems Engineering, Georgia Tech 
Dr. Thomas’ current research projects include the environmental impacts of biofuels and electricity system policy and planning. Her research interests are energy and materials efficiency, sustainability, industrial ecology, technology assessment, international security, and science and technology policy. She has a joint appointment in the School of Public Policy. From 1989 to 2004, she was a Research Scientist at Princeton University, in the Princeton Environmental Institute and in the Center for Energy and Environmental Studies. Thomas was a Member of the U.S. EPA Science Advisory Board from 2003 to 2009. 

Sam Shelton, principal research engineer,Strategic Energy Institute; Georgia Tech
Dr. Shelton’s primary area of teaching, research, and development is innovative energy systems assessment, design and optimization. As its first director, Dr. Shelton founded the Georgia Tech Strategic Energy Initiative in 2004 with a mission to actively engage industry to facilitate short term, high impact energy technology development and commercialization. His current research interests include the use of natural gas for power generation and transportation applications.

The meeting is open to all interested parties, and is free of charge. Refreshments will be served, and there will also be time to network with fellow OR/MS and other professionals in the Atlanta community. Please pass this email along to any associates whom you think might be interested. Please let me know if you plan to attend the meeting and your number of invited guests. I want to try and make sure I have the proper amount of food and drinks. The meeting will be in the Executive Classroom (Room 228) of the ISyE main building located at 755 Ferst Drive. This is the same location as most of our previous meetings. Here's a link to a map of the campus and the ISyE Visitor Parking Lot: http://www.isye.gatech.edu/visitors/maps/isye-map.php. There is plenty, but no longer free, parking. We look forward to seeing you there.

Bio:

Dr. Oburu has over 9 years of analytics and consulting experience with the telecommunication, insurance and retail industries specializing in predictive modeling, CRM analysis, design of experiments and data mining. He is currently a senior statistician with the Strategic Data Analytics group of Equifax. Dr. Oburu was graduated from Georgia State University with a PhD and MA in economics and from Kenyatta University (Nairobi, Kenya) with a BA in economics.

SPEAKER: Dr. Yao Xie, Duke University

ABSTRACT:

Multi-sensor change-point detection, where sensors are distributed to monitor an abrupt emergence of a signal, has attracted much interests recently due to its potential applications in sensor networks, distributed anomaly detection, epidemiology, etc. The emergence of the change-point may abruptly change the distribution of the sensor observations. The goal is to detect such a signal as soon as possible after it occurs, and rarely make false alarm when there is not one. A challenge is that the subset of sensors affected by the change-point is typically unknown and small, i.e., the fraction p of sensors affected is very small. Without taking this signal sparsity into account, we may end up including much noise from the unaffected sensors and hurt detection performance.

We model this sparsity by assuming that each sensor has a probability p0 to be affected by the changepoint, and p0 is a guess for p. Based on this model, we develop a mixture procedure for monitoring parallel streams of data for a change-point that affects only a subset of them, without assuming a spatial structure relating the data streams to one another. Observations are assumed initially to be independent standard normal random variables. After a change-point the observations in a subset of the streams of data have unknown non-zero mean values. The procedure we study uses specific generalized likelihood ratio statistics. An analytic expression is obtained for the average run length (ARL) when there is no change and is shown by simulations to be very accurate. Similarly, an approximation for the expected detection delay after a change-point is also obtained. Numerical examples are given to compare the suggested procedure to other procedures for unstructured problems and in one case where the problem is assumed to have a well defined geometric structure. We discuss sensitivity of the procedure to the assumed value of p0, and extend this to a parallel mixture procedure. Finally, we apply the mixture procedure for real time solar flare detection in videos from the Solar Data Observatory.

Bio:

Dr. Yao Xie received the B.S. degree in Electrical Engineering from the University of Science and Technology of China (USTC), Hefei, China, in 2004, the M.S. degree in Electrical Engineering from the University of Florida, Gainesville, FL, in 2006, and the Ph.D. degree in Electrical Engineering (minor in Mathematics) from Stanford University, Stanford, CA, in 2011. She is currently a postdoctoral Research Scientist with the Electrical and Computer Engineering Department, Duke University, Durham, NC. Her current research interests include statistical sequential methods for signal and information processing, compressed sensing, optimization, and their applications in wireless communications, sensor networks, and medical imaging.

This is an opportunity to have an informal discussion with Dean May. Please note, meetings should be brief so as to accommodate all those seeking to meet with Dean May.

SPEAKER: Professor Yu Zhang,
University of South Florida

ABSTRACT:

This paper presents a study on reliable single and multiple allocation hub-and-spoke network design problems where disruptions at hubs and the resulting hub unavailability can be mitigated by backup hubs and alternative routes. Nonlinear mixed integer programming models and linearized formulations are presented. To solve these difficult problems, Lagrangian relaxation and Branch-and-Bound methods are developed to efficiently obtain optimal solutions. Numerical studies of proposed solution methods on practical instances are reported, along with a few insights of system design.

Bio:

Dr. Yu Zhang is an Assistant Professor with the Department of Civil and Environmental Engineering at the University of South Florida. Dr. Zhang’s research interests include (1) Air traffic flow management; (2) multimodal transportation systems; (3) Air traffic safety; (4) Sustainable transportation planning. Her research projects are funded by government agencies, such as NSF, FAA, FHWA, FDOT and also local industry companies. Dr. Zhang is also conducting international collaboration research with professionals from Europe and East Asia.
Dr. Zhang obtained Ph.D. and Master of Science degrees from the University of California Berkeley. She worked briefly for Jacobs Consultancy (now Leigh Fisher) prior to her joining of USF. Dr. Zhang is the member of two Transportation Research Board (TRB) committees (1) Airfield and Airspace Capacity and Delay (AV060), and (2) Aviation System Planning (AV020). She is the Committee Research Coordinator (CRC) for TRB Av060. Dr. Zhang serves as the Board Director for Chinese Overseas Transportation Association (COTA) and Program Committee for Transportation Research Forum (TRF). Dr. Zhang is the author/co-author of numerous manuscripts that published in prestigious professional journals. She has been invited to serve as the reviewer for journals such as Transportation Science, Transportation Research Part C and the Journal of Transportation Research Record. She also serves as Associate Editor for IEEE Intelligent Transportation Systems Society Conference Management System and Area Editor for COTA International Conference for Transportation Professionals.

Dr. Zhang is the recipient of the 2010 Fred Burggraf Award, Aviation, for excellence in transportation research by researchers 35 years of age or younger, presented by TRB of the National Academies of Science. She received the Best Paper Award in Airline Operations, Quality of Service and Marketing track, in 5th International Conference on Research in Air Transportation held at University of California, Berkeley from May 22- May 24, 201

SPEAKER: Dr. Russell Barton

ABSTRACT:

Discrete-event simulation input model uncertainty affects the process of constructing confidence intervals for the mean response of a system represented by a stochastic simulation. Uncertainty is introduced when input models have been estimated from “real-world” data. The confidence interval should account for both uncertainty about the input models and stochastic noise in the simulation output, but standard practice only accounts for the stochastic noise. Bootstrapping has been used to characterize input model uncertainty, but bootstrap approaches that use simulation replications can be computationally expensive and may fail a requirement for the asymptotic validity of the bootstrap. This talk presents a metamodel-assisted bootstrapping strategy, and compares its performance relative to other approaches for dealing with input uncertainty. This talk presents joint work with Barry Nelson and Wei Xie.

Bio:

Russell Barton is a professor in the Department of Supply Chain and Information Systems at the Pennsylvania State University. He recently completed a two-year assignment as Program Director for Manufacturing Enterprise Systems and Service Enterprise Systems at the U.S. National Science Foundation. Before entering academia, he spent twelve years in industry. He is a past president of the INFORMS Simulation Society and serves on the advisory board for the INFORMS Quality Statistics and Reliability section. He is a senior member of IIE and IEEE. His research interests include applications of statistical and simulation methods to system design and to product design, manufacturing and delivery.

Pizza and drinks will be provided.

2012 SPARKS FORUM FOR ETHICS AND ENGINEERING

...

Is IT Good?

Ethics for an Information Age

 A panel discussion with

...

Casey Fiesler, J.D., Doctoral Student, Human-Centered Computing

Colin Potts, Ph.D., Associate Professor of Interactive Computing and Vice Provost for Undergraduate Education

Valerie Thomas, Ph.D., Associate Professor, Industrial and Systems Engineering and Public Policy

Tuesday, September 18, 2012

4:00 pm - 5:30 pm

* Reception to follow *

Clary Theater, Student Success Center

SPEAKER:  Professor Pierre VANDEKERKHOVE

Maitre de conférences at the department of Analysis and Applied Mathematics of the University Paris-Est Marne-la-Vallée, and invited professor at the department of Materials Science and Engineering, Georgia Institute of Technology

ABSTRACT:

We introduce in this paper a new mixture of regressions model which is a generalization of the semiparametric two-component mixture model studied in Bordes et al. (2006) and Bordes and Vandekerkhove (2010). Namely we consider a two-component mixture of regressions model in which one component is entirely known while the Euclidean parameters and the error distribution of the other component, the mixing ratio, and the distribution of the design data are unknown. Our model is said to be semiparametric in the sense that the probability density function (pdf) of the error involved in the unknown regression model cannot be modeled adequately by using a parametric density family. When the pdf's of the errors involved in each regression model are supposed to be zero-symmetric, we propose an estimator of the various (Euclidean and functional) parameters of the model, and prove under mild conditions its convergence. We prove in particular that, under semiparametric technical conditions all satisfied in the Gaussian case, the Euclidean part of the model is estimated at the rate $o_{a.s}(n^{-1/4+\gamma})$, $\gamma>0$. Finally the implementation and numerical performances of our method are discussed using several simulated datasets and one real microarray dataset (ChipMIX model).

Professor Pierre VANDEKERKHOVE received his Ph.D. in 1998 in Biostatistics at the university Montpellier II, France. His field of research includes Statistics and Applied probability focusing on missing data models and stochastic algorithms, Markov Chain Monte Carlo algorithms.

Contact: "VANDEKERKHOVE Pierre" <Pierre.Vandekerkhove@univ-mlv.fr>

Title: High performance computer algorithms for function approximation and error estimation on arbitrary sparse samples

Abstract:

Stochastic collocation methods are an attractive choice to characterize uncertainty because of their non-intrusive nature. High dimensional stochastic spaces can be approximated well for smooth functions with sparse grids. There has been a focus in extending this approach to non-smooth functions using adaptive sparse grids. We have developed a fast method that can capture piecewise smooth functions in high dimensions with high order and low computational cost. This method can be used for both approximation and error estimation of stochastic simulations where the computations can either be guided or come from a legacy database.

Richard K. Archibald is a staff scientist in the Computational Mathematics group at ORNL. He held the Alston S. Householder Fellowship in Scientific Computing from 2005 until his staff appointment in 2007. Archibald received both his BS in physics and his MS in mathematics from the University of Alberta in Edmonton, Canada in 1996 and 1998, respectively. He obtained his PhD in mathematics from the University of Alberta-Edmonton in 2002.

Archibald¹s current research interests include developing uncertainty quantification algorithms for climate models, designing algorithms for the next generation of high-performance architecture, and establishing long-time integration steps for the dynamical core of climate models at high resolution. He is presently involved in the Center for Advanced Architecture¹s HOMME project, as well as ³Ultra-High Resolution Global Climate Simulations² with principal investigator Jim Hack and researcher Kate Evans. This project seeks to develop the scientific framework to ascertain the benefit of employing very-high-resolution global models to investigate regional-scale phenomena.

This will be a weekend full of fun and excitement where parents and guests will be able to experience what makes Tech a great place to be.

This year, the Yellow Jackets take on the University of Miami Hurricanes for the big game. Game time will be announced on the website 7 days prior to Saturday's game. Click here  for registration information. 

Speaker: Runze Li, The Pennsylvania State University

Abstract:

This paper is concerned with screening features in ultrahigh dimensional data analysis, which has become increasingly important in diverse scientific fields. We develop a sure independence screening procedure based on the distance correlation (DC-SIS, for short). The DC-SIS can be implemented as easily as the sure independence screening procedure based on the Pearson correlation (SIS, for short) proposed by Fan and Lv (2008). However, the DC-SIS can significantly improve the SIS. Fan and Lv (2008) established the sure screening property for the SIS based on linear models, but the sure screening property is valid for the DC-SIS under more general settings including linear models. Furthermore, the implementation of the DC-SIS does not require model specification (e.g., linear model or generalized linear model) for responses or predictors. This is a very appealing property in ultrahigh dimensional data analysis. Moreover, the DC-SIS can be used directly to screen grouped predictor variables and for multivariate response variables. We establish the sure screening property for the DC-SIS, and conduct simulations to examine its finite sample performance. Numerical comparison indicates that the DC-SIS performs much better than the SIS in various models. We also illustrate the DC-SIS through a real data example.

• A unique educational setting with the opportunity to view and “kick the tires” of equipment, systems, and technologies that provide cutting edge solutions to a wide range of supply chain challenges.
• An in-depth look at the parts that make up a world-class supply chain.
• Technical experts who will answer your most difficult questions.
• Special educational sessions seamlessly integrated into the CSCMP education tracks.

To learn more about the annual conference, please visit http://cscmpconference.org/scfuture/overview.asp.

Speaker: Fred J. Hickernell, Illinois Institute of Technology

Abstract:

When: Wednesday, August 15th, at 6:00 PM. 

Where: Executive Classroom of Georgia Tech’s School of Industrial and Systems Engineering.  A reception will precede the meeting at 5:30.

Parking:  There is plenty, but no longer free, parking. Here’s a link to a map of the campus and the ISyE Visitor Parking Lot: http://www.isye.gatech.edu/visitors/maps/isye-map.php.

Dr. Siddhartha Maheshwary (Sid), of Delta Air Lines Inc., will be our guest speaker.  Dr. Maheshwary’s talk will focus on the Crew Reroute and Optimization system used at Delta Airlines. Delta uses this system to reroute the crews in real-time during flight disruptions.

The meeting is open to all interested parties, and is free of charge. Refreshments will be served, and there will also be time to network with fellow OR/MS and other professionals in the Atlanta community.  Please pass this email along to any associates whom you think might be interested.  

The meeting will be in the Executive Classroom (Room 228) of the ISyE main building located at 755 Ferst Drive. This is the same location as most of our previous meetings.  Here’s a link to a map of the campus and the ISyE Visitor Parking Lot: http://www.isye.gatech.edu/visitors/maps/isye-map.php.  There is plenty, but no longer free, parking.

Please RSVP to:

John M. Harris, Ph.D. President INFORMS - Atlanta 404.506.0198

Dr. Siddhartha Maheshwary’s BIO:

Sid got his PhD from the Georgia Tech School of ISyE in 2008 under the advisement of Eva K. Lee. His thesis focused on theory related to cutting planes derived from hypergraph structures found in 0-1 integer programs, and parallel computation for solving integer programming instances. As part of his graduate research, he also developed the RealOpt Simulation-Optimization system used for disaster planning and emergency response preparation. This work received the INFORMS Pierskalla Award in 2005. Upon graduation, he worked for a year at a small startup company in Atlanta where he developed manufacturing and logistics optimization applications. After that he joined the Crew Reroute Group at Delta Airlines, where he is currently working.

SPEAKER: Mark S. Roberts, MD MPP

Professor and Chair, Department of Health Policy and Management

Professor of Medicine, Industrial Engineering and Clinical and Translational Science

University of Pittsburgh Schools of Public Health, Medicine and Engineering

ABSTRACT:

Historically, the application of operations research in health care has focused on the process and delivery of care.  Viewing health care delivery as a production process, operations research and industrial engineering techniques have been use to optimize operating room and ambulance schedules, eliminate bottlenecks in emergency rooms, and re-organize the delivery of radiological services.  There have been some applications in optimizing clinical care; most notable perhaps in the development of algorithms to optimize the delivery of radiation therapy, but these remain rare.

Over the past 15 years, we have been applying methods from operations research to optimize the treatment of disease. The preferred methodology in medicine for acquiring this type of knowledge is the randomized controlled trial.  However, randomized trials are designed to answer simple questions such as “Is A better than B?” when, in fact, most clinical and policy questions are much more complex, and involve picking the best treatment out of a wide array of possibilities, or understanding under what conditions is A better than B. Operations research methods are designed to inform these types of optimization questions, but have been rare in the medical literature.  This talk will describe our efforts to advance and apply operations research techniques to clinical patient care decision and policies, using examples from liver transplantation and HIV care.

Host: Dr. Barbara Boyan, Biomedical Engineering

Presentation Abstract:   GE Energy Monitoring & Diagnostics, based in Atlanta, Georgia, monitors the operation of ~1500 Power Generation units located throughout the world. The M&D center is the “front line” for managing the health of and the information about the GE power generation fleet. To accomplish this task, the M&D team relies heavily on analytics. Today, GE Energy M&D runs over 150 “analytics” continuously on real-time data streaming from the power plant. These “analytics” range from simple rules to complex algorithms, but all are key to how we provide value to GE and our customers. This presentation explores the challenges in managing applied analytics methods in this environment and discusses the significant opportunity that lies ahead.

Mr. Justin Eggart’s BIO:   Justin Eggart is General Manager of Technology for the GE Energy Monitoring and Diagnostics Center in Atlanta, Georgia. Justin and his team are responsible for both operations and technology development for monitoring the health and operation of the GE Thermal power Generation fleet (~1500 units) around the world. The M&D team is the “front line” for operational and maintenance support of the fleet and provides data and analytics services across GE Energy. Justin has over 20 years of experience in developing hardware and software technology for data acquisition applications. Justin joined General Electric in 2002 and has held Engineering, Six Sigma, and Sourcing leadership roles within GE. Prior to joining GE, Justin held various engineering design and management positions at Bently Nevada Corporation including Systems Engineering Leader and Product Line Manager. Justin is a certified Six Sigma Master Black Belt. Justin holds a B.S. Degree in Mechanical Engineering from Montana State University. Justin is active in the community, working with the United Way and GE Volunteers. Justin serves on the Executive Advisory Board for Georgia FIRST and the Industrial Advisory Board for Montana State University. Justin is also an active member of the American Society of Mechanical Engineers.

Georgia Tech is organizing a one-day workshop on the development of sustainable domestic industrial base for lightweight, energy-efficient systems. The workshop will bring together acquisition leaders from U.S. government, academia and industry that have common interests in supply chain analysis and advancing the availability of domestic sources for lightweight material solutions for government systems. 

Additionally, senior leadership and staff representing the interagency Defense Production Act Committee (DPAC) have been invited to participate. The DPAC is a congressionally established body of cabinet secretaries and other agency leaders who are focused on ensuring that the U.S. industrial base can meet government needs. Proceeding of the workshop will be specifically designed to advice the DPAC's newly formed Lightweight Materials Study Group.

Keynote speakers include Hon. Sharon E. Burke, assistant secretary of defense for operational energy plans & programs; Neal Orringer, director of manufacturing at the Department of Defense; Brent Segal, director of research science at the Lockheed Martin Corp. and chief technologist for Lockheed Martin Nanosystems. 

Registration will begin at 7:30 a.m. on June 5. More workshop details coming soon.

Related Links

• 2012 Georgia Tech Interactive Workshop

SPEAKER:  Dr. Bin Wei
Economist, Capital Markets
Board of Governors of the Federal reserve System,
Washington D.C.

ABSTRACT:

We introduce profitability uncertainty into an infinite-horizon variation of the classic Holmstrom and Milgrom (1987) model, and studies optimal dynamic contracting with endogenous learning.  The agent's potential belief manipulation leads to the hidden information problem, which makes incentive provisions intertemporally linked in the optimal contract. We reduce the contracting problem into a dynamic programming problem with one state variable, and characterize the optimal contract with an ordinary differential equation. In the benchmark case of Holmstrom and Milgrom (1987) without learning, the optimal effort is constant, and the optimal contract is linear. In contrast, in our model with endogenous learning, the optimal effort policy becomes history dependent, and decreases over time on average. Moreover, we show that the optimal contract exhibits an option-like feature in that the incentives rise after good performance shocks.

Short bio:
Dr. Bin Wei is currently an economist at the Capital Markets section of Research & Statistics Division at the Board of Governors of the Federal Reserve System. Before joining the Board in September 2011, he had been an assistant professor of finance at Baruch College, the City University of New York for four years between 2007 and  2011. He is interested in a broad range of topics, such as, optimal contracting, delegated portfolio management, liquidity and credit risk modeling, market microstructure. He has published two papers in one of the top finance journals, Review of Financial Studies. Dr. Bin Wei got his Ph.D. in finance from Duke University in 2007 and M.A. in statistics from University of Pennsylvania in 2002.

SPEAKER: Dr. Huairui (Harry) Guo

ABSTRACT:

ReliaSoft is a leading software company providing statistical analysis and project management tools for reliability and quality engineers. We have over 10 major software packages, such as Weibull++ for life data analysis, ALTA for accelerated life test design and analysis, and DOE++ for general experimental designs. In the development of our products, we have encountered many research issues and also received many requirements from our customers. In this talk, I will present several of these issues related to life data analysis and experimental design. I hope we can get input on these matters from experts like you. The possible research topics are:

1)      How to adjust the bias of the maximum likelihood estimates (MLE) of the shape parameter of the Weibull distribution. The bias not only affects reliability prediction, it also affects the standard error of the estimated parameters, which can lead to incorrect conclusions from significance tests such as the likelihood ratio chi-squared test and the approximated standard normal test.

 2)      With censored data, it is sometimes impossible to get the MLE solution for a generalized linear model with the Weibull, exponential or lognormal distribution. How can we determine the existence of the MLE solution for a model by simply examining the data before calling the optimization routine to solve it?

3)      Optimal designs have been widely used in Design of Experiments (DOE). However, with possible censored data from a life test, the traditional optimal design methods that utilize the information matrix cannot be applied directly. Research for designing optimal accelerated life tests with one and two stresses has been conducted, but we would like to explore the possibility of developing a general method for optimal life tests with any number of stresses. All of the current methods are based on the approximated Fisher information bounds (also called Wald confidence bounds); is it possible to design tests based on likelihood ratio bounds, which have been proved more accurate than the Fisher information bounds?

Bio:  Dr. Huairui (Harry) Guo is the Director of Theoretical Development at ReliaSoft Corporation. He received his Ph.D. in Systems and Industrial Engineering from the University of Arizona. His research involves many areas of quality and reliability engineering, including SPC, ANOVA, DOE, repairable and non-repairable system reliability modeling, accelerated life and degradation testing, and warranty prediction. He has been invited to give presentations and seminars for NASA, ASQ, NREL and commercial companies. He has conducted consulting projects for companies from various industries, including renewable energy, oil and gas, automobile, medical devices and semi-conductors. As the leader of the theory team, he is deeply involved in the development of Weibull++, ALTA, DOE++, RGA, BlockSim, Lambda Predict and other products from ReliaSoft. Dr. Guo was the recipient of the Stan Ofsthun Award from the Society of Reliability Engineers (SRE) in 2008 and 2010. He also received the best paper award at the Institute of Industrial Engineers annual research conference in 2007. 

SPEAKER: Professor Shige Peng, Shandong University

ABSTRACT:

A. N. Kolmogorovs Foundations of the Theory of Probability published in 1933, has established the modern axiomatic foundations of probability theory. Since then this theory has been profoundly developed and widely applied to situations where uncertainty cannot be neglected.

But in 1921 Frank Knight has been already clearly classified two types of uncertainties: the first one is for which the probability is known; the second one, now called Knightian uncertainty, is for cases where the probability itself is also uncertain. The situation with Knightian uncertainty has become one of main concerns in the domain of data processing, economics, statistics, and specially in measuring and controlling financial risks. A long time challenging problem is how to establish a theoretical framework, comparable to the Kolmogorovs one, to treat these more general situations with Knightian uncertainties. Motivated from the measure of risk, the objective of the theory of nonlinear expectation rapidly developed in recent years is to solve this problem.

This is a huge program. Some fundamental results have been established and well-understood such as law of large numbers, central limit theorem, martingales, G-Brownian motions, G-martingales and the corresponding stochastic calculus of It?os type, nonlinear Markov processes, as well as the calculation of measures of risk in finance. But many intersting problems are still to be explored, e.g., statistics in the framework nonlinear expectations.

Bio:
Shige Peng is one of the most original and active contributors to the field of probability theory and financial mathematics. Recognized around the world as a leading figure for his work on backward stochastic differential equations and nonlinear expectations, he has had a profound impact in both mathematics and financial engineering. A member of the Chinese Academy of Science, he has been a plenary speaker at the International Congress of Mathematicians, one of the highest honors given to mathematicians.

As a prominent figure in the Chinese mathematics and financial engineering communities he has held numerous visiting professorships and lectured at major institutions and research conferences worldwide, including the Ecole Polytechnique, as well as Osaka, Tokyo, Columbia, Brown, and Princeton universities. He is a Professor First Class at the Institute of Mathematics, Shandong University, and a Distinguished Professor of Ministry of Education of China.

In the course of his visits to Princeton as a Global Scholar in the departments of Mathematics, Operations Research and Financial Engineering, and the Program in Applied and Computational Mathematics, Peng will help nucleate collaborative activities and research interests in the area of stochastic analysis and applications to financial mathematics. He will teach short courses on backward stochastic differential equations and the theory of nonlinear expectations, help organize formal and informal seminars on probability theory and financial mathematics, co- advise undergraduate independent work and graduate students in Mathematics, ORFE and PACM, and collaborate on research with colleagues on campus.

SPEAKER: Professor Ian H. Dinwoodie, Portland State Univ.

ABSTRACT:

An iterative method is given for computing the polynomials that vanish on the basin of attraction of a steady state in discrete polynomial dynamics with finite field coefficients. The algorithm is applied to dynamics of a T cell survival network where it is used to compare transition maps conditional on a basin of attraction.

Contact: Ian Dinwoodie <ihd@pdx.edu>

SPEAKER: Nima Zaerpour, Visiting Scholar from Rotterdam School of Management, Erasmus University

ABSTRACT:

We study temporary storage of fresh produce in a cross-dock center. In order to minimize cooling cost, compact storage systems are used. A major disadvantage of such compact systems is additional retrieval time needed, caused by necessary reshuffles due to the improper storage sequence of unit loads. In practice therefore, a dedicated-storage policy is used in which every storage lane in the system accommodates only one product. However, this policy does not use the planned arrival time information of the products. To make use of this information, this paper proposes a mathematical model considering a shared-storage policy minimizing the total retrieval time. The policy allows different products to share the same lane. In order to solve real-sized problems an effective and efficient heuristic is proposed, based on a greedy construction and an improvement part, which provides near optimal solutions, and which is robust against disturbances in arrival or departure times. The gaps between the results of the heuristic and the lower bound are often zero and mostly less than 1%. Our results, tested for a major distributor of fresh produce, show that the shared-storage policy can reduce the retrieval time by 16% on average as compared to the dedicated-storage policy.

This is a joint work with Dr. John Bartholdi.

Friday, April 20, 2012 Instructional Center (IC) room 105

Bob Foley, Professor in ISyE at Georgia Tech, will be speaking on:

Rare events and Markov chains: a constructive large deviations theory

The talk will be held on Friday, April 20 from 12:00 to 1:00 PM in the Instructional Center (IC) room 105. Lunch will be served before the talk, from 11:45 – 12:00 pm, in the same room. For directions to the venue, please use the following link: http://www.isye.gatech.edu/visitors/maps/ .  

Abstract:   Rare events are events that have small probabilities of occurring. Even though the probabilities are small, rare events can be important. Rare events might correspond to queue lengths exploding or buffers overflowing in a manufacturing system. Even though these "large deviations" are rare, it can be important to know just how rare.

Consider a system that can be modelled as a Markov process and has a stationary distribution that cannot be computed explicitly. We are developing methods for deriving exact asymptotic expressions of the stationary distribution. Furthermore, the methodology gives insight into how various large deviations develop, which allows system designers to know how to improve the system.

During the seminar, I will illustrate the methodology by analyzing a particular queueing system. I will also point out how some of the ideas in this research can be used in rare event simulation.

SPEAKER: Professor Davood Tofighi, GT School of Psychology

ABSTRACT:

Mediational analysis is a statistical approach that examines the effect of treatment (e.g., prevention) on an outcome (e.g., substance use) achieved by targeting and changing one or more intervening variable(s) (e.g., risk and protective factors such as peer drug use norms). Multilevel (mixed model) mediation analysis examines the indirect effect of an independent variable on an outcome achieved by targeting and changing an intervening variable (mediator) in clustered (multilevel) data. We study analytically and through simulation the effects of an omitted variable at Level 2 (cluster level) on a 1-1-1 multilevel mediation model for a randomized experiment conducted within clusters in which the treatment, mediator, and outcome are all measured at Level 1 (individuals). When the residuals in the equations for the mediator and the outcome variables are fully orthogonal, the two methods of calculating the indirect effect (ab, c – c' ) are equivalent at the between- and within-cluster levels. Omitting a variable at Level 2 changes the interpretation of the indirect effect and will induce correlations between the random intercepts or random slopes. The equality of within-cluster ab and c – c' no longer holds. Correlation between random slopes implies that the within-cluster indirect effect is conditional, interpretable at the grand mean level of the omitted variable.

of Industrial and Systems Engineering cordially invites you to attend a celebration for

Professor Ellis Johnson as he retires from

Georgia Institute of Technology

Wednesday the 2nd of May 2012

Speakers: 1:00pm - 4:30pm

Reception: 4:30pm - 6:00pm

Respond by Tuesday the 25th of April

Lisa Basnight

404-385-3888

lisa@isye.gatech.edu

SPEAKER:  Johanna Neslehova

ABSTRACT:

Archimedean copulas are popular dependence structures which may be regarded as extensions of shared multiplicative frailty models, frequently used in biostatistics. This talk is concerned with new rank-based estimators for these copulas. The approach stems from a recent representation of these copulas as the survival copulas of simplex distributions. The procedures are based on a reconstruction of the radial part of the simplex distribution from the so-called Kendall distribution, which arises through the multivariate probability integral transformation of the data. In the bivariate case, the methodology is justified by the well known fact that an Archimedean copula is in one-to-one correspondence with its Kendall distribution. It turns out that this property continues to hold in the trivariate case, and strong evidence is provided that it extends to any dimension. A convenient criterion for the convergence of sequences of multivariate Archimedean copulas will be presented and used to show consistency of the proposed estimators.

Contact: johanna@math.mcgill.ca

SPEAKER: Ramesh Johari

ABSTRACT:

We study learning in a dynamic setting where identical copies of a good are sold over time through a sequence of second price auctions. Each agent in the market has an 'unknown' independent private valuation which determines the distribution of the reward she obtains from the good; for example, in sponsored search settings, advertisers may initially be unsure of the value of a click. Though the induced dynamic game is complex, we simplify analysis of the market using an approximation methodology known as 'mean field equilibrium' (MFE). The methodology assumes that agents optimize only with respect to long run average estimates of the distribution of other players' bids.

We show a remarkable fact: in a mean field equilibrium, the agent has an optimal strategy where she bids truthfully according to a 'conjoint valuation'. The conjoint valuation is the sum of her current expected valuation, together with an overbid amount that is exactly the expected marginal benefit to one additional observation about her true private valuation. Under mild conditions on the model, we show that an MFE exists, and that it is a good approximation to a 'rational' agent's behavior as the number of agents increases. Formally, if every agent except one follows the MFE strategy, then the remaining agent's loss on playing the MFE strategy converges to zero as the number of agents in the market increases.

We conclude by discussing the implications of the auction format and design on the auctioneer's revenue. In particular, we establish a dynamic version of the revenue equivalence theorem, and discuss optimal selection of reserve prices in dynamic auctions.

This is joint work with Krishnamurthy Iyer and Mukund Sundararajan.

Bio:  Ramesh Johari is an Associate Professor at Stanford University and the Cisco Faculty Scholar in the School of Engineering, with a full-time appointment in the Department of Management Science and Engineering (MS&E), and courtesy appointments in the Departments of Computer Science (CS) and Electrical Engineering (EE). He is a member of the Operations Research group in MS&E, the Information Systems Laboratory in EE, and the Institute for Computational and Mathematical Engineering. He received an A.B. in Mathematics from Harvard (1998), a Certificate of Advanced Study in Mathematics from Cambridge (1999), and a Ph.D. in Electrical Engineering and Computer Science from MIT (2004).

He is the recipient of a British Marshall Scholarship (1998), First Place in the INFORMS George E. Nicholson Student Paper Competition (2003), the George M. Sprowls Award for the best doctoral thesis in computer science at MIT (2004), Honorable Mention for the ACM Doctoral Dissertation Award (2004), the Okawa Foundation Research Grant (2005), the MS&E Graduate Teaching Award (2005, 2010), the INFORMS Telecommunications Section Doctoral Dissertation Award (2006), and the NSF CAREER Award (2007). He has served on the program committees of ACM Electronic Commerce (2007, 2009-2011), ACM SIGCOMM (2006, 2011), IEEE Infocom (2007-2011), and ACM SIGMETRICS (2008-2009).

April 19, 2012 3:00pm  Nanotechnology Building tour

                     6:00pm  Dinner

April 20, 2012   8:00am - 3:00pm  Meeting Executive Classroom

Dear ISyE Colleagues,

With the recent publication of Georgia Tech’s Strategic Vision and Plan (Designing The Future) and the ongoing development process for the College of Engineering’s strategic plan, it is timely and appropriate for the School of Industrial & Systems Engineering to likewise re-develop its own strategic plan.  Thus ISyE will be undertaking its own strategic planning process over the next few months, which will yield a plan that shall be congruent with both the Institute’s and College’s visions, yet be uniquely attuned to the aspirations of the School.

The first major step in that process will be an ISyE visioning retreat.  The retreat will enable us to discuss the various aspects of a strategic plan for the School, including its vision, mission, objectives, goals, strategies, and measures.  The retreat will be held on April 30, 2011 at Georgia Tech’s Global Learning Center (more details to follow soon).  You are hereby invited to participate in the retreat.

TITLE:   Adaptive Estimation of Large Covariance Matrices

SPEAKER:  Dr. Ming Yuan

ABSTRACT:

Estimation of large covariance matrices has drawn considerable recent attention and the theoretical focus so far is mainly on developing a minimax theory over a fixed parameter space. In this talk, I shall discuss adaptive covariance matrix estimation where the goal is to construct a single procedure which is minimax rate optimal simultaneously over each parameter space in a large collection. The estimator is constructed by carefully dividing the sample covariance matrix into blocks and then simultaneously estimating the entries in a block by thresholding. I shall also illustrate the use of the technical tools developed in other matrix estimation problems.

Date of the event: April 17, 2012, 5 pm

Location:   ISyE Courtyard (rain location: ISyE Atrium)

SPEAKER:  Jamol Pender, Ph.D. Candidate in ORFE at Princeton

ABSTRACT:

Large scale systems such as customer contact centers, like telephone call centers, as well as healthcare centers, like hospitals, have customer inflow-outflow dynamics with many common features. The customer arrival patterns may have time of day or seasonal effects. Moreover, customer population sizes tend to be large where the individual actions are intrinsic and independent of other customer actions and there are multiple service agents, so many customers have access to services in parallel. Finally, arriving customers engaging in service may be delayed if all the available agents are busy. Moreover, these waiting customers may decide to leave the systems if they feel that their delay in receiving service is excessively long.

A fundamental Markov process, with dynamic rates, queueing model for large service systems is a multiserver queue with non-homogeneous Poisson arrivals as well as service and customer abandonment times that are exponentially distributed. An asymptotic scaling for these queues leads to both functional strong law of large numbers and central limit theorems. The first yields a dynamical system that we call a fluid model. The second scaled limit yields a diffusion process model that is Gaussian under conditions involving the fluid model not lingering too closely to the number of servers. Finally, the fluid model coupled with the mean and covariance of the Gaussian model is also a dynamical system. These results are a special case of a general asymptotic theory for Markovian service networks as derived in the seminal work of Mandelbaum, Massey, and Reiman [1998].

In practice, these results yield useful Gaussian approximations to the transient queue length distribution. However, these estimates tend not to work as well when the lingering effect is significant. We can improve these methods by introducing a new technique that is called the Gaussian-skewness approximation (GSA). It is the special case of a general Hermite polynomial expansion for a Gaussian random variable. We then obtain dynamical systems that approximate the mean, variance and higher cumulant moments of the queueing process for a more accurate, non-Gaussian estimation.

TITLE: Bounding the Performance of Ambulance Deployment Policies

SPEAKER: Shane Henderson

ABSTRACT:

Ambulance organizations everywhere face increasing call volumes, increasing traffic congestion, and shrinking budgets. To keep response times small, many employ some kind of system status management (SSM). SSM is the practice of real-time control of the ambulance fleet, using Global Positioning System (GPS) units on the ambulances to track location, and information from the ambulance crews to track status. Available ambulances are carefully stationed to ensure coverage, while not requiring too many moves of the ambulance crews. I'll review methods for designing SSM policies, and methods for bounding what is achievable with any deployment policy. These bounds are useful in determining whether methods other than SSM need to be employed to achieve contractual targets for response times, and for helping determine when to stop looking for SSM improvements because policies are near optimal. Joint work with Matt Maxwell, Chaoxu Tong and Huseyin Topaloglu.

Bio: Shane G. Henderson is a professor in the School of Operations Research and Information Engineering at Cornell University. He received his PhD from Stanford University in 1997, and has held academic positions in the Department of Industrial and Operations Engineering at the University of Michigan and the Department of Engineering Science at the University of Auckland. His research interests include discrete-event simulation, simulation optimization, and emergency-services planning. He is the current chair of the INFORMS Applied Probability Society, the past simulation area editor for Operations Research, and an associate editor for both Management Science and Stochastic Systems.

SPEAKER: Dr. Phil Kaminsky, UC Berkeley

ABSTRACT:

We consider the capacity investment problem faced by pharmaceutical firms and other firms that have long and risky product research/development/approval cycles for products that require expensive and long lead-time manufacturing capacity in order to meet demand.  These firms must balance two conflicting objectives: on one hand, the delay in scaling-up production once the product is approved must be minimized, and on the other hand, the risk of investing in ultimately unused capacity must be minimized.  In many cases (at least in the pharmaceutical industry), firms are hesitant to reconsider capacity investments once initial investment decisions are made.  To explore alternative strategies, we develop a stylized model of a capacity investment problem where the firm re-evaluates its capacity investment strategy as information about the potential success of the product is updated (for example, via clinical trial results in the case of the pharmaceutical industry).  We characterize optimal investment strategies in a variety of settings, and use a computational study to identify settings in which by frequently reviewing the building strategy, the firm can substantially reduce both the delay of the commercial launch of the new product and the risk of lost investment.    We consider settings where the firm can invest in alternative capacity types with different costs and lead times, explore when the availability of more than one capacity type is most valuable, and investigate whether a firm should ever start to invest in one type of capacity and then switch to building an alternative capacity type.

This research was motivated by projects completed as part of our new Biopharmaceutical Operations Center at UC Berkeley, and I will begin the talk with a brief overview of these efforts.

Joint work with Ming Yuen.    

SPEAKER: Hugh Chipman

ABSTRACT:

Sequential design, or "active learning" can be an effective way to plan a experiment, so as to gain maximal information about a response model. The data-generating mechanism as well as the scientific objective can have important influence on the way in which the design is generated, and the estimated response model. For example, if the objective is to maximize response, we may only be interested in accurate estimates near the maximum. In computer experiments, Gaussian process models are a common approach, and have been used for sequential design and optimization. Instead we use an adaptive nonparametric regression model ("Bayesian Additive Regression Trees", or BART) to deal with nonstationarities and other complex relationships. By providing both point estimates and uncertainty bounds for prediction, BART provides a basis for sequential design criteria to ?nd optima with few function evaluations. Other applications, including sequential design in high-throughput screening for drug discover will also be discussed.

SPEAKER: Franklin Dexter

ABSTRACT:

It’s been 15 years since recognition that planning the operating room hours into which cases are scheduled is a single period (perishable asset) newsvendor problem. A typical (real-world) problem would be choosing between an 8 hr or 10 hr workday given a mean ± standard deviation of workload = 8.6 ± 0.6 hr. The specific application is simple because: (a) there is complete knowledge of the demand and (b) fixed and few different container sizes (e.g., 8 hr and 10 hr days) so that how the optimization is solved has little effect on the quality of the solution. Furthermore, data sources, amount of data to use, and other details are not major issues. Rather, the practical challenges are different. (1) Psychological biases and anesthesia-hospital contracts influence implementation. (2) Lack of use is evident months later when anesthesiologists, nurse anesthetists, and operating room nurses are assigned to cases and there is a mismatch between staff scheduled to work that day with the caseload. 

Bio: During the past 15 years, Franklin Dexter, MD PhD, and his colleagues have developed much of the science in anesthesia group and operating room management. Dr. Dexter completed a Sc.B. in Applied Mathematics & Biology with Honors from Brown University; Masters Degree & Ph.D. in Biomedical Engineering, with specialization in biomathematics, from Case Western Reserve University; M.D. degree from Case Western Reserve University School of Medicine; and Anesthesiology residency at the University of Iowa. He is Professor in the Department of Anesthesia at the University of Iowa. Several times a year, he teaches a four-day intensive course in operating room management. He has given more than more than 135 invited presentations in the United States and abroad. He is Statistical Editor of Anesthesia & Analgesia, as well as Section Editor for Economics, Education, and Policy. Annually he reviews more than 700 papers and grant applications. As Director of the Division of Management Consulting in the Department of Anesthesia, he has performed more than 315 consultations, for more than 130 corporations. He has published more than 325 papers in the fields of operating room management and anesthesia. Details of his background, expertise, contact information, as well as the comprehensive bibliography of scientific papers in OR management, are at www.FranklinDexter.net

SPEAKER: Yuri Nesterov

ABSTRACT:

We consider a new class of huge-scale problems, the problems with sparse subgradients. The most important functions of this type are piece-wise linear. For optimization problems with uniform sparsity of corresponding linear operators, we suggest a very efficient implementation of subgradient iterations, which total cost depends logarithmically in the dimension. This technique is based on a recursive update of the results of matrix/vector products and the values of symmetric functions. It works well, for example, for matrices with few nonzero diagonals and for max-type functions.

We show that the updating technique can be efficiently coupled with the simplest subgradient methods, the unconstrained minimization method by B. Polyak, and the constrained minimization scheme by N. Shor. Similar results can be obtained for a new non- smooth random variant of a coordinate descent scheme. We present also the promising results of preliminary computational experiments.

SPEAKER: Dr. Roshan Vengazhiyil

ABSTRACT:

A new deterministic approximation method for Bayesian computation, knownas Design of Experiments-based Interpolation Technique (DoIt), isproposed. The method works by sampling points from the parameter spaceusing an experimental design and then fitting a kriging model tointerpolate the unnormalized posterior. The approximated posteriordensity is a weighted average of normal densities and therefore, most ofthe posterior quantities can be easily computed. DoIt is a generalcomputing technique which is easy-to-implement and can be applied tomany complex Bayesian problems. Moreover, it does not suffer from thecurse of dimensionality as much as some of the quadrature methods. Itcan work using fewer posterior evaluations, which is a great advantageover the Monte Carlo and Markov chain Monte Carlo methods especiallywhen dealing with computationally expensive posteriors.

Join the Georgia Tech community as we launch the national STAY WITH IT™ campaign. The Day of Engineering is a national pep rally designed to promote and build enthusiasm around STAY WITH IT™, which focuses on helping current engineering students stay in their current field of study, graduate, and begin successful engineering careers. Join Dean Gary May, Intel Corporation President & CEO Paul Otellini, mtvU, Facebook, and other prominent engineering alumni and guests as we celebrate engineering with fun events, music, food, giveaways, and more!

Tech Lounge – 11:30 a.m. – 1:30 p.m.
Tent near Campanile
Rain Location: Student Center Ballroom

Students can drop by the Tech Lounge tent to speak to employment professionals about job search strategies and receive résumé advice. While there, you'll have the opportunity to chat with prominent engineering alumni and special guests from technology and entertainment, as well as receive free prizes for correctly answering engineering trivia. Free drinks and giveaways will be provided, as well as the opportunity to win an Ultrabook™, inspired by Intel®. mtvU will also be nearby filming an upcoming segment showcasing Georgia Tech and the Day of Engineering.

Facebook Live – 2:45 p.m. – 4:00 p.m.
Ferst Center for the Arts

Grab your friends and attend the STAY WITH IT™ Facebook Live panel discussion hosted by Montel Williams, television personality, radio talk show host, actor, and engineer, and featuring Paul Otellini, President & CEO, Intel Corporation, Charles Bolden, NASA Administrator, executives from Facebook, Google, and Comcast along with a special message from President Barack Obama. Be one of the select few to ask the panelists a question that will be streamed to eleven other peer institutions in the U.S. and across the Web. Participants will have the chance to win one of five Ultrabooks. A free ticket is required to attend. Tickets are available from 9:00 a.m. – 4:00 p.m. at the Student Center box office.

Sergiu Hart
Hebrew University of Jerusalem
Game Dynamics and Equilibria

4:30 pm, Klaus 1116, refreshments at 4:00PM in the Atrium

A poster can be downloaded from http://www.aco.gatech.edu/colloq/

The concept of "strategic equilibrium," where each player's strategy isoptimal against those of the other players, was introduced by John Nashin his Ph.D. thesis in 1950. Throughout the years, Nash equilibrium hashad a most significant impact in economics and many other areas.
However, more than 60 years later, its dynamic foundations - how areequilibria reached in long-term interactions - are still not wellestablished.

In this talk we will overview a body of work of the last decade ondynamical systems in multi-player environments. On the one hand, thenatural informational restriction that each participant may not know thepayoffs and utilities of the other participants - "uncoupledness" -
turns out to severely limit the possibilities to converge to Nashequilibria. On the other hand, there are simple adaptive heuristics -such as "regret matching" - that lead in the long run to correlatedequilibria, a concept that embodies full rationality. We will alsomention connections to behavioral and neurobiological studies, tocomputer science concepts, and to engineering applications.

SPEAKER: Dr. Howard Bondell

ABSTRACT:

The triumvirate of outlier resistance, distributional robustness, andefficiency in both small and large samples, constitute the Holy Grailof robust statistics. We show that a two-stage procedure based onan initial robust estimate of scale followed by an application ofgeneralized empirical likelihood comes very close to attaining that goal.The resulting estimators are able to attain full asymptotic efficiency atthe Normal distribution, while simulations point to the ability tomaintain this efficiency down to small sample sizes. Additionally, theestimators are shown to have the maximum attainable finite-samplereplacementbreakdown point, and thus remain stable in the presence of heavy-taileddistributions and outliers. Although previous proposals with fullasymptotic efficiency exist in the literature, their finite sampleefficiency can often be low. The method is discussed in detailfor linear regression, but can be naturally extended to other areas,such as multivariate estimation of location and covariance.

What: This two-day, invitation only Forum will bring together U.S. business, labor, academic and government leaders to discuss how innovations in supply chain and logistics infrastructure and services can stimulate the U.S. Manufacturing global competitiveness and growth.

Registration is complimentary, but seating is limited. Registrants will be informed within 36 hours if seating is available.  Preference will be given to registrants from Council on Competitiveness member organizations.

Who: Sponsored by: The Council on Competitiveness, U.S. Manufacturing Competitiveness Initiative and the Georgia Institute of Technology

SPEAKER: Dr. Wenxuan Zhong

ABSTRACT:

In this talk, a forward screening selection procedure will be discussedunder the sufficient dimension reduction framework, in which the responsevariable is influenced by a subset of predictors through an unknownfunction of a few linear combinations of them. Unlike linear model, ourproposed method does not impose a special form of relationship (such aslinear) between the response variable and the predictor variables. Ourmethod selects variables that attain the maximum correlation between thetransformed response and the linear combination of the variables. Variousasymptotic properties, and in particular, its variable selectionperformance under diverging number of predictors and sample size has beeninvestigated and will be discussed in this talk. The empirical performanceof the procedure will be demonstrated in functional genomic analysis.

Contact: Wenxuan Zhong <wenxuan@illinois.edu>

SPEAKER: Professor Regina Liu

ABSTRACT:

Data depth and its induced center-outward ordering have given rise to many useful tools in nonparametric multivariate analysis. A DD-plot (depth vs depth plot) is the two dimensional scatter plot of depth values of the given sample points with respect to the two underlying distributions. It can be a useful tool to visualize the difference of two distributions. We discuss some of the utilities of DD-plots in this presentation. In particular, we discuss approaches devised from DD-plots to classification (thus named DD classifier) and testing the difference between two samples. The approaches are completely data driven and the classification or test outcomes can be easily visualized on the two-dimensional DD-plot regardless how high the dimension of the data. Moreover, these approaches are easy to implement and they bypasses the task of estimating underlying parameters such as means and scales, often required by the existing statistical approaches. We show that DD-classifier is asymptotically equivalent to the Bayes rule under suitable conditions, and it can achieve Bayes error for a family broader than elliptical distributions. Overall, DD-classifier performs well across a broad range of settings, and compares favorably with existing methods, including KNN and SVM. It can also be robust against outliers or contamination.

This is joint work with Juan Cuesta-Albertos (Universidad de Cantabria, Spain) and Jun Li (University of California, Riverside).

SPEAKER: Professor Weichung Wang

ABSTRACT:

Numerical simulations of three-dimensional photonic crystalshave gathered great importance in the investigations of their physicalproperties and real world applications. A main purpose of thesimulations is to predict particular lattice structures to achievemaximal bandgaps. This target is a challenge as the computations arehighly time-consuming and many tuning parameters are involved. Thesetuning parameters include structure parameters in the photonicscrystal lattices and algorithmic performance parameters in theeigenvalue solvers. To identify the optimal lattice structures and theoptimal algorithm combinations, we use several statistical tools tomodel the unknown response functions of the tuning parameters. Toolsarising in design and analysis of computer experiments will bediscussed to show the details of the schemes. Promising numericalresults will also be presented to justify the proposed schemes.

SPEAKER: Professor Linda Zhao

ABSTRACT:

It is common practice in statistical data analysis to perform data-drivenmodel selection and derive statistical inference from the selected model.Such inference is generally invalid. We propose to produce valid “post-selection inference” by reducing the problem to one of simultaneousinference. We describe the structure of the simultaneous inference problemand give some asymptotic results. We also develop an algorithm fornumerical computation for the width of our new confidence intervals.

This is joint work with Richard Berk, Lawrence Brown, Andreas Buja, and KaiZhang.

SPEAKER: Pengyi Shi

ABSTRACT:

We study patient flow management in an inpatient department of aSingaporean hospital. We focus on understanding the effect of an"early discharge" policy, implemented in late 2009, on wardoverflow rate and on fraction of ED patients who have to wait six hoursor longer to get a bed. We propose a new stochastic network modelwhose service times are not independent, identically distributed (iid)and are dictated by a discharge distribution. We discuss a number ofother key features that need to be built into the model. We show, viaa simulation study, that our model is able to capture the hourlyperformance of the inpatient operation. The model allows one toevaluate the impact of operational policies such as earlydischarge and overflow policies.

This is joint work with Jim Dai(Georgia Tech), Ding Ding (University of International Business &Economics, Beijng), and James Ang and Mabel Chou (NUS).

SPEAKER: Dr. Ben Wang

ABSTRACT:

This talk covers some of Dr. Wang and his colleagues’ recent research anda path forward. The purpose is to seek collaborations with experts instatistics, reliability, quality, process control and optimization.Gen I composites, made with fiberglass and early carbon fibers,served as metal replacements in secondary, non-load bearing parts.The Gen I success fostered the development and progressive use of moderncarbon fibers in load bearing structures to replace metals. Gen IIcomposites, built on intermediate-modulus carbon fibers andimproved matrix resins, brought about a broader use in selected structures.Advancements in the 1980’s-90’s resulted in innovative uses of compositesin commercial aerospace, auto, marine, space and sporting goods andexpanded military applications. This growth was largely driven to achieveadditional weight saving due to the materials’unparalleled abilityto solve seemingly contradictory requirements, such as reducing weighwhile increasing mechanical properties. The potential of composites isclear and the trend of continued use of composites is unstoppable.

Treated as metal replacement, total performance improvements at thesystem’s level has largely been “linear” over the past 50 years spanningGen I and the current Gen II, despite a tremendous body of knowledge inmaterials science and impressive engineering developments.However, if we continue this linear trend and extrapolate into out years,can composites meet much more stringent requirements for tomorrow’slightweight engineered systems characterized by unprecedented requirementsfor performance, quality, energy efficiency, reliability, safety,environmental compatibility and life cycle affordability?

We define Generation III composite systems as ultra-lightweight, energyefficient, high-performance composite structures where multiple functionsco-exist symbiotically without requiring parasitic components. Such aparadigm-changing endeavor obviously requires enormous multidisciplinaryteamwork over a long period of time.

The presentation is intended to stimulate a discussion on the followingtopics:

1. Can synergistic materials with intrinsic properties be developed, scaled and integrated to realize effective Gen III multifunctional structural systems of the highest possible quality, reliability and performance throughout their life cycles?

2. Can the Gen III multifunctional structural systems exceed the performance of today’s best composites and if so, by how much?

3. What are the barriers, challenges and possible solutions at the basic science and enabling technology levels and how can these solutions be embodied in an integrated engineered system? 

SPEAKER: Dr. Chang W. Kang

ABSTRACT:

In some manufacturing processes, the variance can change, depending onthe process mean. For example, if the output value reflects processyield, an increased mean might naturally lead to an increase in variance.When the variance is a function of the mean, the coefficient of variation(CV) is an appropriate measure for process variability. Since the CVcontrol chart was first introduced by Kang et al (2007), there were sometrials to improve the performance of CV control charts depending on theshift size.

In this research, we present some CV related control charts and comparethe performance of those control charts for better use in the field.The CV control chart shows good sensitivity to large shift in CV.The CV-EWMA control chart(2008), the CV-DEWMA control chart(2011)and the CV-GWMA control chart(2011) were developed to controlprocesses sensitively responding to small shifts of CV.The FIR CV-GWMA control chart is more effective control chartto detecting off-target processes in the stage of set-up or start-up ofprocess.

Host: Professor Paul Kvam <paul.kvam@isye.gatech.edu>

SPEAKER: Sebastian Pokutta

ABSTRACT:

Limits of system performance can often understood in the context of information and entropy. In the talk we give two examples. One theoretical one where a strong bound on the size of a smallest possible representation is obtained via an entropy argument. The second deals with an application where the goal is to eliminate information asymmetries using optimization methods.

More precisely:

In the first part, we solve a 20-year old problem posed by M. Yannakakis and prove that there exists no polynomial-size linear program (LP) whose associated polytope projects to the traveling salesman polytope, even if the LP is not required to be symmetric. Moreover, we prove that this holds also for the maximum cut problem and the stable set problem.

In the second part, we consider a real-world energy market coupling problem which aims for a more balanced and consistent determination of prices in adjacent markets in presence of coupling mechanisms. By doing so the amount of possible arbitrage is minimized.

(The first part is joined work with: Samuel Fiorini, Serge Massar, Hans Raj Tiwary, and Ronald de Wolf // the second part is joined work with: Alexander Martin and Johannes Müller)

SPEAKER: Dr. Russell Barton
Program Director, Manufacturing Enterprise Systems and Service Enterprise Systems
National Science Foundation

ABSTRACT:

This talk will provide a broad survey of research areas supported by the Manufacturing Enterprise Systems (MES) and Service Enterprise Systems (SES) programs at NSF. It will also cover a number of NSF-wide initiatives related to systems engineering issues in health care and manufacturing that are also important for prospective proposers, and present a crystal-ball view of possible future initiatives. The talk will conclude with a characterization of winning and losing strategies for proposal writing for the MES and SES programs.

Biography

Russell Barton is Program Director for manufacturing management and services management research in the MES and SES programs. These areas have a combined annual budget of over $9 million. He is on assignment at NSF from the Smeal College of Business at Penn State, where he is a professor in the Department of Supply Chain and Information Systems. He previously served as co-director for the Penn State Master of Manufacturing Management degree program, and as associate dean for research and Ph.D./M.S. programs for the Smeal College. He holds a B.S. in electrical engineering from Princeton University and M.S. and Ph.D. degrees in operations research from Cornell University.

SPEAKER: Dr Elisabeth Pate-Cornell

ABSTRACT:

A presentation of four models of risk and decision analyses involving games, designed to support strategic decisions. The first is a single-move game, in which the US faces risks of terrorist attacks by several possible groups. The result is a probabilistic ranking of the threat posed by different types of weapons.  The second is a dynamic counter-terrorism analysis designed to compare the stabilizing effects of different short- and long-term government strategies. It is based on the simulation of an alternate game between a government and a terrorist group. The third is a dynamic evaluation of US nuclear counter-proliferation strategies for a single country. It involves an analysis of the weapon development program given the country's evolving intent and capabilities, and the effectiveness of different US strategies to prevent or delay its success.  The fourth is a principal-agent model of the development of an engineered system, in which an agent in charge of part of the project may consider meeting a deadline by cutting corners if he falls behind schedule. This generally increases the system failure probability and the goal is to support the decisions of the manager in setting constraints and incentives to decrease the total costs. These four cases are based on systems analysis and probability but present a spectrum of models, assumptions and results.

Biography:

Dr. M. Elisabeth Paté-Cornell was born in Dakar, Senegal, in 1948. Her undergraduate degree is in mathematics and physics (BS, Marseilles, France, 1968), and her first graduate degrees are in applied mathematics and computer science (MS and Engineer Degree, Institut Polytechnique de Grenoble, France, 1970; 1971).  She received a Masters degree in Operations Research (OR) in 1972 and a Ph.D. in Engineering-Economic Systems (EES) in 1978, both from Stanford University.  She joined the Stanford faculty in 1981, and serves as Chair (since 1997) of the now Department of Management Science and Engineering.  In 1999, she was named the Burt and Deedee McMurtry Professor in the School of Engineering, and she is also a Senior Fellow (by courtesy) of the Stanford Institute for International Studies.

Dr. Paté-Cornell was elected to the National Academy of Engineering in 1995, and is currently a member of its Council. She has served on the President’s Foreign Intelligence Advisory Board from December 2001 to December 2004.  Dr. Paté-Cornell is a world leader in research related to engineering risk analysis, risk management, decision analysis under uncertainty, and more generally, the use of Bayesian probability to process incomplete information. In recent years, her research and that of her Engineering Risk Research Group at Stanford have focused on the inclusion of both technical and organizational factors in probabilistic risk analysis models. These models have been applied to a wide variety of topics, ranging from the risk management of the NASA shuttle tiles to that of offshore oil platforms and medical systems such as anesthesia during surgery. She is currently working on risk management processes for complex projects and programs, with application to space, industrial and medical systems. Since 2001, she has applied risk analytic methods to the study of different types of terrorist attacks on the United States, the assessment of intelligence information and the effectiveness of counter measures.

Dr. Paté-Cornell is the author or co-author of more than a hundred papers in refereed journals and conference proceedings. She has received several best-paper awards from professional organizations such as the American Nuclear Society and the Decision Analysis Society of INFORMS (for her work on the shuttle tiles), and peer-reviewed journals such as Military Operations Research in 2002 for a paper on the assessment of terrorist threats.

India is one of the world's most dynamic countries and is designated as a priority market by the U.S. Department of Commerce. India's economy grew on average 8 percent over the last ten years and growing dynamism in several of its regional markets have created wide and diverse business prospects for U.S. companies. The country's supply chains are responsible for its continued growth and globalization. The sector is valued by official studies at $500 billion.

India's infrastructure needs present an unprecedented opportunity for U.S. firms in developing supply chains, warehouses, cold chain storage, ports, roads, railroads, airports, and power grids. India will invest $1 trillion in its infrastructure during the 12th Five-Year Plan (2012-2017) and looks for private sector participation to fund half of this massive expansion through the Public-Private Partnership (PPP) model. India's highly entrepreneurial and rapidly globalizing private sector is investing in infrastructure projects and is actively looking for suitable partners in USA.

Who Should Attend?

This one-day global business forum is designed to be intensive and specific for business decision-makers, strategic business planners, export, marketing and operations managers, top professionals, legal practitioners, consultants, and academics with a special interest in India's Supply Chain and the business impacts of its continuing growth and globalization. We suggest sending a team of executives to the Forum.

Forum's Major Emphases:

• Identifying market opportunities in India for U.S.-based firms in various promising sectors
• Where, when, and how to build your India market
• Market entry and joint venture strategies
• Overcoming regulatory and cultural obstacles

Plus! - Candid Industry and Business Case Panels with experienced executives.

Organized By

The USA India Business Summit (UIBS), the Georgia Tech Center for International Business Education & Research (CIBER) and Georgia College & State University.

With Support of

The U.S. Commercial Service, The Georgia Department of Economic Development (GDEcD), The Metro Atlanta Chamber, Confederation of Indian Industry (CII), Am Cham India, Georgia Tech Supply Chain and Logistics Institute and MM Activ

Sponsored by

Other ISCMOC sponsors include USIBRC, Georgia College, Infosys, Manhattan Associates, DACHSER Intelligent Logistics, Metro Atlanta Chamber and USIII.

Program

8:30-9:15am Registration and Continental Breakfast
9:15-9:30am Welcoming Remarks

• Ani Agnihotri, Program Chair, India's Supply Chain: Markets and Opportunities Conference
• Ms. Gretchen Corbin, Deputy Commissioner, Global Commerce , Georgia Department of Economic Development
• Honorable Ajit Kumar, Consul General, Indian Consulate, Atlanta

9:30-10:15am Keynote Address - India's Integration with Global and USA Supply Chain

• Mr. L. Radhakrishnan, I.A.S., Chairman, Jawaharlal Nehru Port Trust, Mumbai, India
• Dr. John McIntyre, Prof. of Intl. Business and Director, GaTech, CIBER - Session Moderator

10:15-11:00am Morning Special Session - A Perspective on US and Georgia’s Supply Chain & Logistics Industry

• Mr. Page Siplon, Executive Director, Center of Innovation for Logistics, Georgia Department of Economic Development
• Mr. Tom Singh, Vice President, Global Supply Chain, CIRCOR International, Inc. - Session Moderator

Coffee Break

11:15-12:15pm Panel A - Building Supply Chain and Logistics Infrastructure in India

Ports, Airports and other opportunities

• Mr. David R. Gault, Director Business Development India, DACHSER USA
• Mr. Steve Mulaik, Partner, The Progress Group
• Mr. Warren Jones, Aviation Development Manager, Hartsfield Jackson Atlanta Intl Airport

11:15-12:15pm Panel B - India opportunities in Logistics Sector

Warehousing and Cold Chain, Freight forwarder and Air Freight

• Ms. Jaymie C. Forrest, Managing Director, Georgia Tech Supply Chain and Logistics Institute – Session Moderator
• Mr. J.D. Schwefler, Vice President, International Business Development, Americold
• Mr. Chris Barnes, President, Warehouse Education and Research Council Atlanta Chapter
• Mr. Jayant Bharadwaj, President, Swift Freight (USA) Inc

12:30-2:00pm Luncheon Keynote Address - Opportunities and Challenges in India's Supply Chain and Logistics Sector

• Dr. S.B. Agnihotri*, I.A.S., Director General of Shipping, India
• Mr. Thomas Strauss, Southern Regional Director, U.S. Department of Commerce – Session Moderator

2:15-3:00pm Afternoon Special Session - Challenges and Opportunities for Global Multinationals: How New Business Models, Scale, Technology and Frugal InnovationCan Combine to Create Real Breakthroughs in India's Supply Chains

• Mr. Sandeep Dadlani, VP & Head Americas, Retail, CPG and Logistics, Infosys, India
• Dr. Matthew Liao-Troth, Dean, J. Whitney Bunting College of Business, Georgia College & State Univ. - Session Moderator

Coffee Break

3:15-4:15pm Panel C - Indian Supply Chain and Logistics Sector

Agriculture and Food, Mergers & Acquisitions and Business Partnerships

• Mr. Bret Douglas, Vice President, Global Trade & Supply Chain, HSBC Bank
• Mr. Dharmendra Gangrade, Head – Logistics, Pidilite Industries Ltd., India
• Dr. Douglas F. Britton, Manager, Agricultural Technology Research Program, Food Processing Technology Division, GTRI

3:15-4:15pm Panel D - IT and Other Services

IT and Allied Technologies, Enterprise Resource Planning and Warehouse Management Systems

• Mr. Vivek Sehgal, Senior Director, Manhattan Associates
• Mr. Michael Stolarczyk, President, Kontane Logistics

4:15-5:00pm Case Studies - Case studies of Warehousing and distribution in India

Food, Retail and Technology

• Mr. Bill Bastian II, President, Bastian Solutions

5:15-7:30pm Reception

Conference Fees:

• Early Registration: $175/person (subject to availability)
• Registration: $225/person

Registration Link: http://www.usaindiabusinesssummit.com/registration.php

SPEAKER: Professor Wolfgang Haerdle

ABSTRACT:

On the temperature derivative market, modeling temperature volatility isan important issue for pricing and hedging. In order to apply pricingtools of financial mathematics, one needs to isolate a Gaussian riskfactor. A conventional model for temperature dynamics is a stochasticmodel with seasonality and intertemporal autocorrelation.Empirical work based on seasonality and autocorrelation correction revealsthat the obtained residuals are heteroscedastic with a periodic pattern.The object of this research is to estimate this heteroscedastic functionso that after scale normalisation a pure standardised Gaussian variableappears. Earlier work investigated this temperature risk in differentlocations and showed that neither parametric component functions nor alocal linear smoother with constant smoothing parameter are flexibleenough to generally describe the volatility process well. Therefore, weconsider a local adaptive modeling approach to find at each time point, anoptimal smoothing parameter to locally estimate the seasonality andvolatility. Our approach provides a more flexible and accurate fittingprocedure of localised temperature risk process by achievingexcellent normal risk factors.

Contact: "Wolfgang Haerdle" wolfgang.k.haerdle@me.com

SPEAKER:  Dr. Lee Dicker

ABSTRACT:

Motivated by questions about dense (non-sparse) signals inhigh-dimensional data analysis, we study the unconditional out-of-sampleprediction error (predictive risk) associated with three classes of denseestimators for high-dimensional linear models: Ridge regressionestimators, scalar multiples of the ordinary least squares estimator(which we refer to as James-Stein estimators), and marginal regressionestimators. Our results require no assumptions about sparsity and implythat in linear models where the number of predictors is roughlyproportional to the number of observations:(i) If the population predictor covariance is known (or if anorm-consistent estimator is available), then the ridge estimatoroutperforms the James-Stein estimator; (ii) both the ridge and James-Steinestimators outperform the ordinary least squares estimator, and theimprovements offered by these estimators are especially significant whenthe signal-to-noise ratio is small; and (iii) the marginal estimator hasserious deficiencies for out-of-sample prediction. We derive newclosed-form expressions for the asymptotic predictive risk of theestimators, which allow us to precisely quantify the previous claims.Additionally, minimax ridge and James-Stein estimators are identified.Finally, we argue that the ridge estimator is, in fact, asymptoticallyoptimal among dense estimators for out-of-sample prediction inhigh-dimensional linear models.

Contact: Lee Dicker <ldicker@stat.rutgers.edu>

SPEAKER: Rong Duan

ABSTRACT:

Learning and identifying events in network traffic is crucial for serviceproviders to improve their mobility network performance. In fact, largespecial events attract cell phone users to relative small areas, whichcauses sudden surge in network traffic. To handle such increased load,it is necessary to measure the increased network traffic and quantifythe impact of the events, so that relevant resources can be optimizedto enhance the network capability. However, this problemis challenging due to several issues: (1) Multiple periodic temporaltraffic patterns (i.e., nonhomogeneous process) even for normal traffic;(2) Irregularly distributed spatial neighbor information;(3) Different temporal patterns driven by different events even forspatial neighborhoods; (4) Large scale data set.

This paper proposes a systematic event detection method that dealswith the above problems. With the additivity property of Poisson process,we propose an algorithm to integrate spatial information by aggregatingthe behavior of temporal data under various areas. Markov ModulatedNonhomogeneous Poisson Process (MMNHPP) is employed to estimate theprobability with which event happens, when and where the events takeplace, and assess the spatial and temporal impacts of the events.Localized events are then ranked globally for prioritizing moresignificant events. Synthetic data are generated to illustrate ourprocedure and validate the performance. An industrial example from atelecommunication company is also presented toshow the effectiveness of the proposed method.

Contact: rongduan@research.att.com

SPEAKER: Prof. Jan van Vuuren, Univ. of Stellenbosch, South Africa

ABSTRACT:

 In this talk we consider the Evolutionary Spatial Prisoners' Dilemma (ESPD) in which players are modelled by the vertices of an underlying graph G representing some spatial organisational structure amongst the players. During each round of the ESPD every pair of adjacent players in G play a classical prisoners' dilemma against each other, and they update their strategies from one round to the next based on the perceived success (as measured by pay-off values) achieved by the strategies of their neighbours during the previous round. In this way players are able to adapt and learn good strategies from each other as the game progresses, without understanding why these strategies are good. We characterise all steady states of the ESPD for the two cases where G is a path or a cycle, and we also characterise those initial states that lead to the emergence of persistent substates of cooperation over time. We finally determine analytically (i.e. without using simulation) the probability that the game's states will evolve from a randomly generated initial state towards a steady state which accommodates some form of persistent cooperation.

Joint work with Alewyn Burger & Martijn van der Merwe   

SPEAKER: Marty Puterman

ABSTRACT:

This talkwill highlight research carried out by the CIHR (Canadian Institutes of HealthResearch) Team on Operations in Quality Cancer Care which I head up.  In this talk, I will focus on two verydifferent scheduling applications, one applied study which uses multi-criteriadiscrete optimization to schedule chemotherapy patient’s daily appointments anda more fundamental study which uses approximate dynamic programming todetermine effective patient scheduling rules for radiotherapy treatments. A brief description of each topic follows:

1. Chemotherapy AppointmentScheduling Process Redesign: Manual bookingpractices in place at the start of this study limited effective demandmanagement and resulted in last-minute rescheduling ofappointments.  The consequence of thiswas heightened stress for patients and staff and operational challenges for thepharmacy and outpatient clinics. The implementation in June 2011 of moreflexible booking procedures combined with a custom-built computerizedscheduling program based on a multi-criteria discrete optimization model, hasalleviated these problems by providing a reasonable timeframe to notifypatients of their appointments.  This hasreduced unnecessary changes to pre-booked appointments, supportedthe complex task of organizing the daily treatment schedule andbalanced nurse and pharmacy workload. 

2. Dynamic Radiotherapy (RT) Appointment Scheduling:  This research sought to develop good policiesfor the dynamic scheduling of patients for radiation therapy.  A unique feature of this problem is thatscheduling a patient means committing capacity over a course of treatments thatcan range from 1 to 28 days depending on cancer site and treatmentprotocol.  Further patients differ withrespect to the degree of urgency for their treatment and which RT machines candeliver their therapy.  The practicalproblem motivating the research involved scheduling 11,000 patients per year on9 RT machines. To address it, we formulated and solved a discountedinfinite-horizon Markov decision process (MDP). We used an affine architectureto approximate the MDP value function and solved an equivalent linearprogramming model through column generation to obtain an approximate optimal policyfor this problem. The benefits of the proposed method are evaluated by simulatingits performance for a practical example based on data provided by the BCCA inwhich relative costs of delays were assessed by RT professionals.  We hope this research will provide the basisfor development of a scheduling application.

Co-authorsinclude Antoine Saure, Jonathan Patrick, Scott Tyldesley, John French, PabloSantibanez, Ruben Aristizabal, Vincent Chow, Kevin Huang and Nancy Runzer

Bio:

MartinL. Puterman is Advisory Board Professor of Operations in UBC’s Sauder School ofBusiness.  He was founder and director ofthe Centre for Operations Excellence (in Sauder), the UBC Centre for HealthCare Management,  and the BiostatisticalConsulting Service at BC Children’s Hospital. He is co-principal investigatorof the CIHR Team for Operations Research in Quality Cancer Care.

 His research focuses on health care operationsresearch especially pertaining to cancer care delivery and decision making,Markov decision processes and statistical modeling of golf performance andPGAtour structure.  He has consultedwidely on health care operations, statistical modeling, inventory control,forecasting, operations management, program evaluation and management strategy.

He received the prestigious INFORMSLanchester Prize for his book MarkovDecision Processes.    He is anINFORMS Fellow and recipient of the Canadian Operations Research Society (CORS)Award of Merit, the CORS Practice Prize and the INFORMS case prize.   He has been an editorial board member of Mathematics of Operations Research, Operations Research, Management Science,Production and Operations Management, Manufacturing and Service OperationsManagement and The Journal of the American StatisticalAssociation.

Hereceived his PhD in Operations Research and an MS in Statistics from StanfordUniversity and AB in Mathematics from Cornell.

SPEAKER: Brian Denton

ABSTRACT:

Prostatecancer is the most common solid tumor that affects American men. Screeningtypically involves the use of prostate specific antigen (PSA) tests. However,the imperfect nature of PSA tests, and the potential for subsequent harm fromunnecessary biopsies and treatment, has raised debate about whether and when toscreen. In this talk I will provide some background on prostate cancer, currentscreening guidelines, and a summary of the recent controversy over PSA testing.Next, I will discuss a partially observable Markov decision process (POMDP)model to investigate the optimal design of screening policies. Screening policies are defined by the patient’s probability of havingprostate cancer which is estimated from their history of PSA tests resultsusing Bayesian updating. The core states are the patients’ prostate cancerrelated health states. Transition probabilities among health states areestimated using a large longitudinal dataset from Olmsted County, the MayoClinic Radical Prostatectomy Registry (MCRPR) and the medical literature.  Reward functions that are considered includequality adjusted survival (patient perspective) and costs (third party payerperspective).

Sometheoretical properties that define the optimal policy will be discussed, and anew approximation method suited to solving finite horizon non-stationary POMDPswill be presented.  Theresults of computational experiments will be used toillustrate the use of the model for making screening decisions, such as if andwhen to recommend a patient for a PSA test, and when to refer patients forbiopsy and subsequent treatment.  Sensitivityanalysis will be presented to demonstrate the relative importance of factorsthat define patient specific preferences and risk factors. Finally, futureresearch directions in the area will be discussed.  

SPEAKER:  Sergei Savin

ABSTRACT:

We study the impact of physician workload on hospital reimbursement utilizing a detailed data set from the trauma department of a major urban hospital. We find that the proportion of patients assigned a  "high-severity'' status for reimbursement purposes, which maps, on average, to a 27.8% higher payment for the hospital, is substantially reduced as the workload of the discharging physician increases. This effect persists after we control for a number of systematic differences in patient characteristics, condition and time of discharge. Furthermore, we show  that it is not caused by selection bias or endogeneity in either discharge timing or allocation of discharges to physicians. Finally, we find that the impact of workload on the probability that a patient is assigned the high-severity designation is moderated by the department's past experience in handling certain patient conditions. We attribute this phenomenon to a workload-induced reduction in diligence of paperwork execution. We estimate the associated monetary loss to be approximately 0.9% (with a 95% confidence interval of 0.3% and 1.5%) of the department's annual revenue.

Speaker:  Diego Morán, ISyE - Georgia Tech

Abstract:

A mixed integer point is a vector in $\mathbb{R}^n$ whose first $n_1$ coordinates are integer. We present necessary and sufficient conditions for the convex hull of mixed integer points contained in a general convex set to be closed. This leads to useful results for special classes of convex sets such as pointed cones and strictly convex sets. 

Furthermore, by using these results, we show that there exists a polynomial time algorithm to check the closedness of the convex hull of the mixed integer points contained in the feasible region of a second order conic programming problem, for the special case this region is defined by just one Lorentz cone and one rational matrix.

This is joint work with Santanu Dey.

Pizza and refreshments will be served.

• The Impact of The Panama Canal Expansion
• Leveraging Canada
• Collaboration in Mexico
• International Freight
  
• Global Trade and Country Trade/Logistics Reviews forCanada, Mexico, China, Panama, Singapore and India

The forum will include speakers from Worley Parsons Group, CSX, Florida East Coast Railway,Transport Canada, Transplace Mexico, DB Schenker Logistics, Home Depot,Wal-Mart and Carters. Also included is a joint-session with theAtlanta CSCMP Chapter on the afternoon of May 1st which will provide a richnetworking opportunity!

ABOUT THE SCEF
Georgia Tech's Supply Chain Executive Forum (SCEF), an unit of the Supply Chain & Logistics Institute, represents THE most relevant and valuable opportunity for senior supply chain executives to enhance the strategic impact of their supply chain processes and activities. The SCEF meets twice each year at Georgia Tech in Atlanta and helps its members to identify new and compelling ways to streamline operations to enhance profitability, integrate supply chain strategy with corporate strategy, and grow professionally within and beyond their current organizations.

To learn more about Georgia Tech’s Supply Chain Executive Forum, visit http://www.scl.gatech.edu/scef and read about our Spring 2011 and Fall 2011 meetings.

Where: Hamburg, Germany at the Curio-Haus

Who: The conference, which was initiated and organized by the Georgia Tech Health and Humanitarian Logistics Center since 2009, will be hosted by The Kühne Logistics University (KLU) in Hamburg, Germany, at the Curio-Haus. The conference is co-organized by the KLU-INSEAD Research Center on Humanitarian Logistics, the Georgia Tech Health and Humanitarian Logistics Center, and the Humanitarian Logistics Association (HLA).
What: The focus of this year's conference is "Creating Sustainable Health and Humanitarian Systems" and brings together high level speakers from across the health and humanitarian sectors, including non-governmental organizations (NGOs), industry, government, etc.

For more information about the conference program, please visit http://humlog2012.the-klu.org/ or http://www.humanitarian.gatech.edu/humlog2012/ or email humlog2012@the-klu.org or humlog@isye.gatech.edu.