BIOGRAPHY
Dr. Maria Vlasiou is currently a Research Engineer II in the Stewart School of ISyE. She received her Ph.D. in Mathematics from the Eindhoven University of Technology, The Netherlands, and holds a Mathematics degree from the Aristotle University of Thessaloniki, Greece. During the 2001-2002 academic year, she worked at the University of Bielefeld in Germany.

Dr. Vlasiou's research interests are in stochastic processes and stochastic operations research with applications in manufacturing, call centers, warehousing, healthcare logistics, and traffic control. Her honors include four scholarships from the Greek State Scholarships Foundation and a five-year grant for post-graduate studies.

Open to the general public.

Panelists/Speakers:

Paul Raines, The Home Depot; Executive Vice President of U.S. Stores

Steve Hollingworth, CARE; Chief Operating Officer

Dr. Julie Swann, H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech; Assistant Professor and Co-Director of the Center for Humanitarian Logistics

Moderator:

Dr. Chelsea C. White, III, H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech; School Chair and Professor

Schedule:

6:00

Dr. Pulleyblank will speak on "Computing, Business, and Operations Research: The Next Challenges."

Lecture Abstract
There have been two consistent drivers over the last sixty years of the evolution of computing: Computer power and price/performance improve by a factor of two every eighteen months; the problems that we wish to solve require this growth in capability and more.

We seem to be reaching inflection points with both of these drivers. High performance systems are turning to massive parallelism to continue the required growth in performance. New challenges are arising in business and industry that require the solution of fundamentally different problems as well as the development of new approaches to old problems. Moreover, the rapid growth of a global economy has given an unprecedented urgency to dealing with these changes.

I will review these subjects and some approaches that are being applied, with varying degrees of success. In particular, I will discuss five technical problems that must be solved to enable us to successfully meet the business challenges that we will face in the future.

Biography of William R. Pulleyblank
William R. Pulleyblank, IBM Vice President, is the leader of the Center for Business Optimization within IBM Business Consulting Services. This group develops and deploys high powered optimization and analytic capabilities to improve the business performance of a broad range of companies.

Prior to this, he was the Director of Exploratory Server Systems and Director of the Deep Computing Institute within IBM Research. These research teams provided broad-based support to IBM

Logistics is now looked upon as a key corporate leverage point for simultaneously increasing shareholder and customer value. Unfortunately, the formal links between logistics performance, cost, and value are not yet well defined.

In "Logistics Performance, Cost, & Value Measures" we describe and demonstrate how improving logistics performance in customer service, fill rate, response time, inventory turns, receivables, transportation efficiencies, and warehousing productivity translate into operating and capital cost reductions and increased shareholder value.

We define the link between logistics performance, cost, and shareholder value; define a full and complementary set of logistics performance benchmarks; and demonstrate how to use logistics performance gap analysis to justify new logistics initiatives.

Who Shold Attend:
The course is for professionals who need to better understand how to measure and improve their financial, productivity, quality, and cycle time performance.

How You Will Benefit:
* Learn the Characteristics of Effective Performance Measurement Systems
* Define Metrics
* Set Effective Targets
* Setup a Logistics Performance Measurement Program
* Design Supply Chain Scoreboards
* Define a Metrics Implementation Plan
* Define Data Requirements
* Implement Logistics Perfomance Measures
* Benchmark Logistics Performance
* Tie Logistics performance to Compensation Plans
* Develop Balanced and Comprehensive Logistics Performance Measures
* Use Logistics Performance Measures for Project Justification
* Measure the financial, productivity, quality, and cycle time performance for supply chains, logistics operations, customer service, inventory management, procurement, transportation, and warehousing.

PROGRAM DESCRIPTION
This program focuses on effective planning and management of highly integrated product and information flows from the supplier

Title: Selections

Abstract: In this talk, I will commence by discussing a general paradigm of selections. Subsequently, I will brief my current research before moving on to address regularization parameter selection issues for the SCAD estimator of regression models.

Bio: Professor Tsai is a recognized expert in the practical application of statistics in business, including regression analysis, model selection, high dimensional data, time series, biostatistics, application of statistics in business. He has had more than 80 research papers published in the academic journals relating to statistics, marketing, finance and biostatistics. He teaches courses on forecasting and managerial research methods, and time series analysis and forecasting.

Tsai has developed new statistical models based upon regression and time series analysis, and co-authored a book on the subject. He has collaborated with other Graduate School of Management faculty to produce a new method to analyze the effectiveness of advertising, which helps companies and media buyers determine an unbiased optimal advertising budget. He has also worked on projects to develop statistical methods that can be applied when analyzing long-term investment strategies. He has applied one of the newest techniques in statistics, called Sliced Inverse Regression, which helps predict sales response to individual customers based on where they live, their age, size of family, income, zip code, and other factors.

Speaker: Bernardo K. Pagnoncelli

I will give a brief introduction to chance-constrained programming showing an example and motivating the subject. Then I will present a chance-constrained portfolio problem with random returns. Under normality, the problem can be efficiently solved, but in more general situations we have to rely on approximations. We apply different approximation schemes to the portfolio problem and we present the computational results. The talk will be aimed at a broad audience and no previous knowledge of chance-constrained programming is required.
_______________________________

Bernardo K. Pagnoncelli has an B.S. and a M.Sc. in Mathematics from the Pontifical Catholic University of Rio de Janeiro, Brazil, and is a Ph.D. candidate in Mathematics at the same institution.

He was invited by Professor Shabbir Ahmed to spend one year as a Research Scholar at ISyE, working on Stochastic Programming.

The Maximum Flow Problem is a fundamental problem in operations research. This talk focuses on two problems involving computing a maximum flow. First, we present theoretical results on the Maximum Flow Network Interdiction Problem (MFNIP), where an interdictor seeks to remove arcs from a network so as to minimize the maximum flow in the remaining network. Second, we introduce an algorithm to rapidly solve an online sequence of similar maximum flow problems. As an application, we extend our algorithm to compute a robust minimum s-t cut, that is, a conservatively chosen minimum s-t cut in light of data uncertainty. We report the substantial reduction in running time from using our algorithm instead of repeatedly using an efficient black-box maximum flow solver.

About Dr. Valerie Thomas:
Most academics specialize in either public policy or engineering. Georgia Tech Associate Professor Valerie Thomas explores the nexus between the two. Thomas, who holds a joint appointment in the School of Public Policy and the H. Milton Stewart School of Industrial & Systems Engineering, sees public policy and sustainable technology as complementary fields. "Dealing with the future energy problems of the world is difficult and the problems are fundamental engineering problems," explains Thomas, who held postdoctoral appointments in nuclear arms control at Carnegie Mellon and Princeton and spent last year working with Congress on key environmental issues as an American Physical Society Congressional Science Fellow. To work on the world's energy problems, she researches a variety of engineering technologies, from making recycling and waste management more efficient using information technology to reduced fossil fuel use and air transportation. She also attends weekly Environmentally Conscience Design and Manufacturing Program meetings to discuss research on environmentally benign products. "We're working together and building links," she says of the professors and students who attend the meetings. "There's a lot of expertise at Georgia Tech that doesn't exist elsewhere."

Title: Copula Based Independent Component Analysis

Abstract: We propose a parametric version of Independent Component Analysis (ICA) via Copulas-families of multivariate distributions that join univariate margins to multivariate distributions. Our procedure exploits the role for copula models in information theory and in measures of association, specifically: the use of copulae densities as parametric mutual information, and as measures of association on the rank statistics.

The purpose of the Spring Research Conference on Statistics in Industry and Technology is to promote research in statistical methods that address problems in industry and technology. This focus is to be interpreted broadly to cover a wide range of application areas including manufacturing and information sciences. The conference is intended to stimulate interactions among statisticians, researchers in the application areas, and industrial practitioners. It will provide a forum where participants can describe current research, identify important problems and areas of application, and formulate future research directions.

Featured Plenary Speakers for the 2008 Spring Research Conference include:

Peter Bickel
Professor of Statistics
University of California, Berkeley

James Glimm
Distinguished Professor
Dept of Applied Mathematics and Statistics
University at Stony Brook

Zhong Lin Wang
Regents' Professor and
Director, Center for Nanostructure Characterization
Georgia Tech

Organizing Committee:
Xiaoming Huo, Vladimir Koltchinskii, Justin Romberg, Ming Yuan

Sponsored by:
National Science Foundation
School of ECE
School of Math
School of ISyE, Georgia Tech

Guest Lecturer: Prof. Steve Kou
Industrial Engineering and Operations Research
Columbia University

Presentation Abstract:
Choosing a proper risk measure is of great regulatory importance and is relevant to the interfaces of operations and finance, as exemplified in Basel Accord which uses VaR (or quantiles) in combination with scenario analysis as a preferred risk measure for banking and operational risk. Two main families of axiomatically based risk measures are the coherent risk measures, which assume subadditivity for random variables, and the insurance risk measures, which assume additivity for comonotonic random variables. We propose new, data-based, risk measures, called textit{natural risk statistics}, that are characterized by a new set of axioms. The new axioms only require subadditivity for comonotonic random variables, consistent with the prospect theory. We point out while many risk measures may be suitable for internal risk management, robustness is an important consideration for external risk measures. Comparing to the previous measures, the natural risk statistics include the tail conditional median which is more robust than the tail conditional expectation suggested by coherent risk measures; and, unlike insurance risk measures, the natural risk statistics can also incorporate scenario analysis. The natural risk statistics includes VaR as a special case and therefore shows that VaR, though simple, is not irrational.

(Joint work with Chris Heyde and Xian Hua Peng)

Any integer program may be relaxed to a group problem.We define the master cyclic group problem and several master knapsack problems, show the relationship between the problems, and give several classes of facet-defining inequalities for each problem, as well as a set of mappings that take facets from one type of master polyhedra to another.

ABSTRACT:
With the rapid growth of the Internet, network link speeds have become faster every year to accommodate more Internet users. Measuring and monitoring the traffic on such high-speed links has become an ever challenging problem. Our team is among the first to realize that data streaming algorithms have the potential to become very useful tools for measuring and monitoring high-speed networks. Data streaming is concerned with processing a long stream of data items in one pass using a small working memory in order to estimate certain statistics of the stream. The challenge is to use this small memory to "remember" as much information pertinent to this estimation as possible. However, existing data streaming algorithms are typically designed for processing a single stream of data for a single type of statistic, and most of these algorithms cannot operate at very high link speed. Targeting the deficiency of existing approaches, we have investigated new data streaming paradigms and mechanisms that allow us to perform large-scale distributed data streaming on tens of thousands of high-speed links and nodes, and aggregate, compress, and interpret these streaming results, for better measurement and monitoring of large networks. We also discovered that the applications of data streaming go far beyond network measurement and monitoring. Recently, we have successfully applied the data streaming techniques to a seemingly unrelated area: query routing in an unstructured P2P network. In this talk, I will talk about these new applications as well as our theory of formulating the data streaming problems in CS as the channel design problem in EE.

BIOGRAPHY:
Jun (Jim) Xu is an Associate Professor in the College of Computing at Georgia Institute of Technology. He received his Ph.D. in Computer and Information Science from The Ohio State University in 2000. His current research interests include data streaming algorithms for the measurement and monitoring of computer networks, algorithms and data structures for computer networks, network security, and performance modeling and simulation. He received the NSF CAREER award in 2003 for his ongoing efforts in establishing fundamental lower bound and tradeoff results in networking. He is a co-author of a paper that won the Best Student Paper Award from 2004 ACM Sigmetrics/IFIP Performance joint conference, and the faculty advisor of the student winners. He received 2006 IBM faculty award for making fundamental contributions to performance evaluation methodologies.

February 19

Masakiyo Miyazawa 2:00 to 2:40
David McDonald 2:40 to 3:20
Break 3:20 to 3:50
Sunil Kumar 3:50 to 4:30

February 20

Wuqin Lin 9:00 to 9:40
Baris Ata 9:40 to 10:20
Break 10:20 to 10:50
Hayriye Ayhan 10:50 to 11:30

All talks will take place in room 228 (Executive classroom) of the main building. The titles and the abstracts are given below.

Masakiyo Miyazawa, Tokyo University of Sciences

Answers to the tail decay rate problem for a reflected random walk on a positive quadrant

This talk is concerned with a skip free random walk on the two dimensional positive quadrant with homogeneous reflecting transitions at each boundary face. We are interested in the tail decay behavior of its marginal distributions in the steady state. We solve this problem by characterizing the decay rates as the solutions of a certain non linear optimization problem, provided the stationary distribution exists. In general, those decay rates are in the logarithmic sense, referred to as weak decay rates. However, this result with help of existence results also answers to the question when the asymptotic decay is exactly geometric and what happens if it is not the case. We exemplify these results for Jackson networks and their modifications.

============================================================

David McDonald, University of Ottawa

A Stochastic Model for the Throughput of HTTP sessions

We introduce a simplified model of an HTTP which consists of a succession of idle and download periods. The file downloads are subject to a fixed packet loss probability. The same TCP connection is used for the download of a random number of files. We assume the files are transmitted in the congestion avoidance regime and we don't consider time outs. We do take into account the effect of slow start at the beginning of a download period with SSTHRESH set to CWND after the last loss in the preceeding download period; i.e. the cycles are dependent. For this stochastic model, we derive a closed form formula for the stationary throughput obtained by a flow. We also derive closed form expressions for the mean time to transfer a file and for the distribution of the throughput. We also briefly discuss how the formulas can be applied to predict bandwidth sharing among competing HTTP flows.

============================================================

Sunil Kumar, Stanford University.

Dynamics of New Product Introduction in Closed Rental Systems

We study a rental system where a fixed number of heterogeneous users rent one product at a time from a collection of re-usable products. The online DVD rental firm Netflix provides the motivation. We assume that rental durations of each user are i.i.d. with finite mean. We study transient behavior in this system following the introduction of a new product that is desired by all the users. We represent the usage process for this new product in terms of an empirical distribution. This allows us to characterize the asymptotic behavior of the usage process as the number of users increases without bound, via appropriate versions of Glivenko-Cantelli and Donsker

In the second part, we propose a two-factor equilibrium model for electricity spot and futures prices. Not only does our model capture features such as spikes and seasonality, but it also has some other properties. First, it can incorporate oligopoly. Second, the spot prices have infinite expectations, but the futures prices have finite expectations.

Cai's Research Interests
Financial Engineering
-Asset pricing, electricity derivatives pricing, computational finance
-Jump diffusion models for asset pricing, modeling electricity market
-Numerical inversion of Laplace transform
Applied Probability in Operations Research and Finance

Education
2008 (Expected) Ph.D. in Operations Research, Columbia University, New York, NY
Advisor: Steven Kou
2005 M.S. in Operations Research, Columbia University, New York, NY
2003 M.S. in Probability and Statistics, Peking University, Beijing, P. R. China
2000 B.S. in Probability and Statistics, Peking University, Beijing, P. R. China

Guest Lecturer: Professor William Li
Operations and Management Science Department
University of Minnesota

Presentation Abstract: We discuss the problem of designing an experiment for selecting a good model from a set of models of interest. The research is built on the work on model-robust design of Li and Nachtsheim (2000) and model-discriminating designs of Jones, Li, Nachtsheim, and Ye (2007, 2008). We introduce new criteria for model discrimination and use these and existing criteria to evaluate standard orthogonal designs. We also use these criteria to construct optimal two-level designs for screening experiments. Results indicate that, for a given sample size and number of desired factors, not all orthogonal designs are model-discriminating designs for the model spaces considered. We conclude with some simulation studies, which show that the proposed designs can lead to a higher probability of identifying the correct model in the data analysis procedure than traditional minimum aberration designs.

Presentation Title: Service-level differentiation in large-scale service systems

Abstract: Motivated by telephone call centers, we study the problem of service-level differentiation in largescale service systems with multiple customer classes and multiple agent pools. Call Centers usually specify their operational objectives through the x-y service-level (SL) constraint, which stipulates that x% of the calls should be answered within y seconds. Typically, service-level differentiation is imposed through different SL objectives for different customer classes. In order to achieve these constraints a Call-Center needs to determine the types of agents that will be used (Design), the number of agents of each type (Staffing) and the real-time control to be used (Routing). We propose a solution based on a family of routing controls called Fixed-Queue-Ratio (FQR) rules. In many settings, FQR allows us to solve the combined Design-Staffing-and-Routing problem in a nearly optimal way. Our analysis also establishes a diminishing-returns property of flexibility by showing that,under FQR, very moderate cross-training is sufficient to make the call center work as efficiently as a single-pool system. This is joint work with Ward Whitt.

Education
2004-2008 Graduate School of Business, Columbia University, New York.
PhD in Decisions, Risk and Operations
Dissertation: "Staffing and Control of Many-Server Service Systems". Principal Adviser: Ward Whitt; Co-Advisers: Mor Armony, Costis Maglaras and Assaf Zeevi.

2002-2004 Technion

Presentation Title: Contact Tracing to Control Endemic Infectious Disease

Abstract:
Contact tracing (also known as partner notification) is a disease control strategy where the contacts of infected individuals are sought and tested. Contact tracing is the primary means of disease control for low-prevalence infectious diseases, but its cost-effectiveness has not been modeled. My research merges traditional epidemic control models (as epidemiologists might develop) with OR concepts of resource allocation and economic analysis, and develops models analyzing the effectiveness and cost-effectiveness of contact tracing. I use differential equation and network simulation models to estimate the costs and effectiveness of various screening and contact tracing strategies, and I use optimization to determine the optimal disease control strategy (for example, the optimal allocation of resources between screening and contact tracing).

Event Contacts: Paul Griffin (and Julie Swann)

Title: Data Fusion for Quality Improvements in Complex Systems

Abstract:
Our current data rich environment provides unprecedented opportunities and research challenges for improving complex system operations both in manufacturing and in service industries. The development of a novel data fusion methodology is currently very relevant, and will provide a long lasting future impact. Efforts in data fusion research have significantly advanced the development of methodology in areas like rapid change detection, root cause diagnostics, and system control. Data fusion, through integrating engineering domain knowledge with data analysis techniques from advanced statistics, signal processing, decision making, and system control, represents one of the frontiers in quality improvement research for complex systems.

This presentation will provide an overview of ongoing data fusion research for manufacturing system improvement and its extension to service industry applications. The basic concepts in data fusion research will be discussed, with emphasis placed on promoting the integration of disparate methodologies into a cohesive entity to enable effective decision-making for system operations. Examples of methodological developments and their applications will be discussed in order to demonstrate the characteristics of data fusion research and to emphasize the need for multidisciplinary system integration efforts. Detailed discussions will focus on an example of a model-free multiscale process monitoring method for autocorrelated processes.

Contact person: Dave Goldsman (or Jennifer Harris)

Lecturer: Mark L. Spearman, Factory Physics, Inc.

Abstract:
In recent years, companies have begun to realize the severity of the risk facing the entire corporation by not addressing uncertainty in the supply chain. Indeed, a recent survey indicated that more than two-thirds of the companies surveyed had experienced a supply chain disruption from which they took more than one week to recover. Furthermore, the study revealed that 73 percent of the executives surveyed had a major disruption in the past five years. Of these, 36 percent took more than one month to recover.

Why does it take companies so long to recover?
Why are so many companies susceptible to disruption?
Do the new "Advanced Planning and Optimization" systems offer any help?

I believe there are at least two reasons why modern supply chains are so susceptible to such disruption: (1) supply chains are not designed with uncertainty in mind and (2) modern planning and scheduling systems and practices are not robust enough to operate under conditions significantly different from those for which they are planned. I will address these and other issues in this talk and also offer a solutionknown as Dynamic Risk-based Scheduling (DRS).

Dynamic Risk-based Scheduling avoids the need to produce a detailed schedule that becomes obsolete as soon as any one of the assumptions changes. Instead of producing a schedule, DRS calculates scheduling parameters that then are used to dynamically determine the schedule. Moreover, DRS provides a method to incorporate recourse strategies in a practical way. For instance, if an alarm indicates there is insufficient capacity, extra capacity can be scheduled before any orders are late. However, deciding how much extra capacity to have and how often it should be used in a difficult optimization problem. The result is a system that is both more robust and more effective than other advanced scheduling techniques.

Presentation Title: Primal-dual interior-point methods with asymmetric barriers

Abstract:
Very often, in the standard optimization problem with general cone constraints, the natural self-concordant barriers for primal and dual cones are not self-conjugate. Geometric programming and the power cones provide us with the most important examples. This primal-dual asymmetry destroys the good properties of the primal-dual barrier function. As a result, the standard machinery of the long-step infeasible-start primal-dual methods does not work.
In this talk, we show that even in the asymmetric case, the interior-point methods remain a powerful computational tool. We discuss several potential-reduction and path following primal-dual schemes. We show that some of them can be implemented in a matrix-free way. This opens a possibility for a direct competition of these polynomial-time methods with recently revived fast gradient schemes, which are applicable to very large optimization problems.

(parts of the talk are based on joint work with Yi Lin and Vladimir Koltchinskii respectively.)

Speaker:
Chris M. Schlegel
Manager, Enterprise Risk Management Group
Southern Company

Abstract:
Schlegel will provide a brief discussion of recent modeling efforts of the Risk Analysis Services group. The group was recently formed to model and address certain commodity and interest rate risks within the context of the Enterprise Risk Management group at Southern Company. Applications to structured products in Libor markets, hedging using weather derivatives, and emissions trading will be demonstrated as well.

Speaker: Professor Srinagesh Gavirneni, Cornell University

Abstract: We study a supply chain with one supplier and many retailers that face exogenous end-customer demands. In contrast to the retailers' newsvendor type ordering behavior, we propose a scheduled Synchronized Ordering Policy (SOP), under which all the retailers order freely in the same and only one period of an ordering cycle, and receive fixed shipments in other periods. We identify mathematical conditions under which SOP outperforms the newsvendor type ordering and corroborate them with numerical studies. We also compare SOP to Balanced Ordering Policy (BOP), under which the retailers take turns to order freely in one period of the cycle, and receive fixed shipments in other periods.

Abstract: We consider the Hamiltonian cycle problem (HCP) embedded in a controlled Markov decision process. In this setting, HCP reduces to an optimization problem on a set of Markov chains corresponding to a given graph. We prove that Hamiltonian cycles are minimizers for the trace of the fundamental matrix on a set of all stochastic transition matrices. In case of doubly stochastic matrices with symmetric linear perturbation, we show that Hamiltonian cycles minimize a diagonal element of a fundamental matrix for all admissible values of the perturbation parameter. In contrast to the previous work on this topic, our arguments are primarily based on probabilistic rather than algebraic methods.

Joint work with Vladimir Ejov

Bio: Nelly Litvak is an Assistant Professor at the University of Twente, The Netherlands, in the Stochastic Operations Research group. Her research interests are in stochastic networks, popularity measures and probabilistic ranking schemes in complex networks, queueing theory, perturbed Markov chains, queueing models in health care logistics, performance analysis of carousel systems.

For more information, visit: http://wwwhome.math.utwente.nl/~litvakn/

This talk presents LP and IP formulations that are used to answer a small open question in voting theory. In a much studied election scheme, every voter arranges a set of candidates in order of preference. The outcome of the election is determined by comparing the fraction of the total voters preferring u to v, for each pair of candidates u and v. By treating the candidates as vertices and including a directed edge uv whenever candidate u is preferred to candidate v by a prespecified fraction p of the voters, every election outcome can be represented as a directed graph. It is relatively easy to show that every directed graph represents a possible election outcome for some p>

There will be many opportunities to speak with students and professionals about the work environment and activities in the field of quantitative finance. The symposium is attended by students from the Georgia Tech MS QCF Program and students from educational programs strongly overlapping with the QCF area; by researchers and practitioners working within the QCF area who want to share their experiences and learn more from others in the QCF area; and by other individuals seeking information about the content of the very diverse QCF area.

To register or learn more about the program visit http://www.qcf.gatech.edu/qcfday08/.

Title: Perspectives of Supply Chain in China, and Survey Results

Abstract:
In this talk, Prof. Lin will first provide an overview of logistics and the evolvement of distribution systems in China. He will compare the logistics infrastructure and performance among Asia Pacific region, and point out various bottlenecks in production, consolidation, customs and transportation. Finally, he will present the survey results from users and providers of logistics services. The survey results include inventory turns, costs, material handling in both production and retail segments. The survey also provides an insight into market demand and provider capacities. Prof. Lin will finally share some of his own views in the future of the logistics in China.

Title: An Analysis of Monthly Mean Temperature Trends in the United States

Abstract: To study monthly mean temperature trends for the continental United States, we develop a periodic linear regression model with multiple changepoints, which allows for a mean shift at each changepoint time and simultaneously estimates both trend and changepoint effects. The error component in the model is allowed to be a periodically correlated time series. The model is applied to over 900 temperature recording stations across the U.S., and a spatially smoothed map of temperature trends is presented. Overall, the results suggest more warming than is estimated in previous studies.

http://www2.isye.gatech.edu/statistics/index.php

Speaker: Juan Pablo Vielma

Abstract: Many combinatorial constraints over continuous variables such as SOS1 and SOS2 constraints can be interpreted as disjunctive constraints that restrict the variables to lie in the union of m specially structured polyhedra. Known mixed integer binary formulations for these constraints have a number of binary variables and extra constraints that is linear in m. We give sufficient conditions for constructing formulations for these constraints with a number of binary variables and extra constraints that is logarithmic in m. Using these conditions we introduce the first mixed integer binary formulations for SOS1 and SOS2 constraints that use a number of binary variables and extra constraints that is logarithmic in the number of continuous variables.We also introduce the first mixed integer binary formulations for piecewise linear functions of one and two variables that use a number of binary variables and extra constraints that is logarithmic in the number of linear pieces of the functions. We prove that the new formulations for piecewise linear functions have favorable tightness properties and present computational results showing that they can significantly outperform other mixed integer binary formulations. Joint work with Prof. George L. Nemhauser

Contact: ISyE DOS Optimization Seminars http://www2.isye.gatech.edu/dos/

Title: Control of Complex Integrated Automated Systems

Stefano Barone is assistant professor of Statistics for experimental and technological research at the University of Palermo, Italy. He is moving as associate professor at the Chalmers University of Technology, Sweden

Background
Human beings develop and produce increasingly complex systems constituted of many interrelated and interacting components. Human beings are themselves such a kind of complex systems.

During their life, complex systems are unavoidably subject to failures. Failures are random events, their randomness due to: a) constructive imperfections, b) partly unpredictable conditions of operation, and c) natural phenomena of degradation.

The random nature of failures has stimulated researchers during the last decades to study methods for predicting and assessing the reliability, and methods for developing robust systems, i.e. as much insensitive as possible to the unavoidable variation affecting them during their life cycle.

Particularly important in the so called

Title: Two Problems on the Intersection of Revenue Management and Statistics

Abstract: In this presentation, Dr. Zeevi will describe two problems (well, maybe closer to one & a half...)that lie at the heart of revenue management practices, and are addressed by suitably blending ideas from operation research and statistics. The first is testing the validity of a demand model, and the second is pricing under changing market conditions. The analysis and proposed solutions rely on ideas from hypothesis testing and non-parametric statistics, and change-point detection and sequential analysis, respectively. These problems illustrate possible roles of statistical techniques in revenue management applications, and concurrently indicate how revenue management considerations motivate the development of new statistical methods.

In this paper, we propose and analyze a new approach to finding robust portfolios for asset allocation problems. It differs from the usual worst case approach in that a (dynamic) portfolio is evaluated not only by its performance when there is an adversarial opponent (``nature"), but also by its performance relative to a fully informed benchmark investor who behaves optimally given complete knowledge of the model (i.e. nature's decision). This relative performance approach has several important properties: (i) optimal decisions are less pessimistic than portfolios obtained from the usual worst case approach, (ii) the dynamic problem reduces to a convex static optimization problem under reasonable choices of the benchmark portfolio for important classes of models including ambiguous jump-diffusions, and (iii) this static problem is dual to a Bayesian version of a single period asset allocation problem where the prior on the unknown parameters (for the dual problem) correspond to the Lagrange multipliers in this duality relationship.

Abstract: Localization, mapping, and goal-directed search are three fundamental robot motion problems. In localization, the robot knows the map of the terrain, but does not know its own location. In mapping, the robot must construct the map by traversing the terrain. In goal-directed search, the robot knows the location of its destination relative to its initial location, but does not know the intervening terrain. I'll survey and derive various algorithms and performance bounds for these problems. This is joint work with Apurva Mudgal, Sven Koenig, Joe Mitchell, and possibly others as time permits.

Contact: DOS Optimization Seminars(http://www2.isye.gatech.edu/dos/)

Joint work with J. Foreman (MIT), J. Alspaugh (Dell), R. Bhatnagar (NTU), C.
Teo (NTU), C. Dubois (ENSMP)

Abstract: Dell's supply chain for desktops involves Asian vendors shipping components by sea to several U.S. plants. While suppliers are responsible for shipping enough inventory, Dell can reroute and expedite their shipments while in transit to balance inventory across sites and dynamically track geographic demand patterns. We present the optimization models developed in collaboration with Dell to address this routing challenge, insights gained since their implementation last year including the results of an impact assessment study, and ongoing extensions.

Bio: J

In conjunction with Developing a Design/Simulation Framework
A Workshop with CPDA's Design/Simulation Council
May 13, 2008 ($495.00)

For more information or to register, visit https://cpd-associates.com/download/index.cfm?download=08Atlanta&company=

Faculty Candidate: Mustafa Akan, Kellogg School of Management, Northwestern University

Mustafa Akan, who is working with Baris Ata at Kellogg, will discuss his research interests in in stochastics, revenue management and probably even health (in this area he has worked on a liver transplant problem).

Presentation Title: Robust Operational Aircraft Routing

Abstract:
We developed an approach to ensure that a schedule is maintenance and operationally friendly. The results of fleet assignment models (FAM) often do not meet airlines

Title: Extreme Inequalities for Multiple-Constraint Group Relaxations of Mixed Integer Programs

Abstract:
One of the most successful cutting planes used in Mixed Integer Programming (MIP) solvers, the Gomory Mixed Integer Cut (GMIC), is derived using single constraint relaxation of a MIP. It is, in fact, an extreme inequality for the single-constraint infinite-group relaxation of a MIP. Numerical and theoretical studies suggest that

Abstract:
The past several months have been an eventful period for the U.S. financial markets, mainly due to the crisis in subprime credit markets and the difficulty in modeling collateralized debt obligations (CDOs). In this talk we will present two of financial engineering problems that are related to the recent financial turmoil: (1) How to model CDOs by incorporating clustering defaults? We propose a new model based on Polya processes and the cumulative intensity of counting processes that can calibration the current CDO data very well. (2) What are good external risk measures for the financial regulators? We propose a new axiomatic approach to justify the current governmental regulations, by using axioms from robust statistics and the separation by hyperplanes from functional analysis.

Dr. Audrey Lee is an economist in the Office of Policy and International Affairs at the U.S. Department of Energy, where she began as a Presidential Management Fellow. Her work focuses on economic and policy analysis of energy and climate change policies, using a variety of models and covering an extensive array of technologies. Recent work has involved global biofuels supply, clean coal electricity generation with carbon capture, renewable portfolio standards, and clean energy production tax credits to address climate change. In 2007, Dr. Lee wrote sections of the /World Energy Outlook: China and India Insights/ at the International Energy Agency. Her work there included modeling energy use in the Chinese industrial sector and coastal provinces and the energy embodied in China's exports. Dr. Lee's experience also includes internships at the Federation of American Scientists and Intel Corporation. Dr. Lee earned a B.S. in applied physics from the California Institute of Technology and a M.A. and Ph.D. in electrical engineering from Princeton University.

Presentation Title: A Sample Path Approach to Stochastic Fluid Flow Models

Abstract: Recently, in a series of papers Soohan Ahn and I provided a sample path based approach to the analysis of Markov modulated fluid flow models yielding extremely fast, stable and accurate algorithms for the computation of both steady state and transient results in both the finite and infinite buffer cases. The approach based on ideas that parallel the matrix-geoemetric method in the discrete case can be considered as a stochastic discretization of the fluid model. Current research of mine are focused on generalizing these as an approach to stochastic integration and to apply it to the interesting area mathematical finance.

Title: "Lessons learned from the development of a simulation-based scheduling tool"

Abstract: Over the last two years, we developed a workforce scheduling solution forcomplex assembly lines. The main objective was to make optimum use of the available workers while maximizing the number of jobs processed during a given time horizon. The workers have a large range of qualifications, the jobs have due dates, and there are several resource modes for most of the productions steps (for instance, short processing times with large number of workers or smaller processing time with fewer workers). Typical manufacturing environments where these conditions can be found are the assembly of planes, turbines, printing machines, or other large one-of-a-kind assembly outputs. Due to the huge number of constraints on scheduling in these environments, it is hard even to find a feasible schedule for theworkers. This was the main reason why we decided to start the development of a simulation-based scheduler. We are more than half way through thisdevelopment process and there are several important "lessons learned" with respect to the solution approach as such and the data and information requirements we require from our partners in industry. In this talk, I will discuss modeling issues (e.g., is it possible to develop a common model description for a variety of industries?), solution approach issues (e.g., is it possible to find a feasible solution on time?), and data/information collection issues (e.g., which tools do we need to obtain relevant data/information from our partners in industry). I will not come with answers to all the questions but we have at least some opinions due to our project experiences.

Andy Dishner's talk will focus on answering four questions:
How do I know if I should Outsource all, some or none?
How do I assess my company

SPEAKER: Manoj Chair

ABSTRACT:

We will introduce optimization functionality available in the SAS system through the modeling language SAS procedure OPTMODEL. Using illustrative examples, we will introduce the algebraic modeling features of OPTMODEL as well as the wide array of powerful new solvers for linear, mixed integer linear and non linear programming that it has direct access to. We will also demonstrate that OPTMODEL provides an intuitive high level programming environment which, when combined with the solvers, can be used to implement application specific customized algorithms and heuristics.

Speaker Bio:

Manoj Chari is Director of the Operations Research R&D department at SAS Institute. Before joining SAS in 2003, he was a member of the Mathematics faculty at Louisiana State University, Baton Rouge, where he spent over a decade conducting research and teaching in combinatorics and operations research. He has a Ph.D. in Operations Research from UNC Chapel Hill and was a postdoctoral fellow in Combinatorics and Optimization at the University of Waterloo, Canada. He has also held visiting positions at the Technical University of Berlin and the Indian Institute of Science.

SPEAKER: Dr. Shabbir Ahmed

ABSTRACT:

We report polyhedral results for a class of combinatorial optimization problems involving submodular utility maximization. Strong valid inequalities are derived by lifting lower dimensional inequalities. We provide a complete characterization of the lifting function and its subadditive approximation. Computational experiments with capital budgeting problems show the effectiveness of the derived inequalities within a branch-and-cut approach.

Joint work with Alper Atamturk, UC Berkeley

Paper available at http://www.optimization-online.org/DB_HTML/2008/04/1948.html

Contact: ISyE DOS Optimization Seminars (http://www2.isye.gatech.edu/dos/)

For more information, visit:
http://www.scl.gatech.edu/professional-education/scef/

SPEAKER: Dr. Cerry M. Klein, Program Director, Service Enterprise Systems, Manufacturing Enterprise Systems, National Science Foundation

ABSTRACT:

Service now accounts for approximately 80% of the economy and manufacturing has held steady at 20%. Research in understanding and promoting these two key areas is essential to the continued growth of our economy and the understanding of how these sectors distinctly operate, interact and compliment each other. This presentation will discuss the efforts and funding opportunities at the National Science Foundation in these areas. Additionally, as part of the presentation an open discussion related to the future of manufacturing and service will be held.

SPEAKER: Dr. George Moustakides

ABSTRACT:

In binary hypothesis testing, when hypotheses are composite or the corresponding data pdfs contain unknown parameters, the generalized likelihood ratio test (GLRT) constitutes a popular means for deciding between the two possibilities. GLRT has the very interesting characteristic of performing simultaneous detection and estimation in the case of parameterized pdfs or combined detection and isolation in the case of composite hypotheses. Even though this test is known for years and has been the decision tool in numerous applications, existing results demonstrate only large sample size asymptotic optimality.

In our presentation we introduce a novel, finite sample size detection/estimation formulation for the problem of hypothesis testing with unknown parameters and a corresponding detection/isolation setup for the case of composite hypotheses. The optimum test that results from our performance measure optimization has a GLRT-like structure which is closely related to the criterion we employ for the parameter estimation or isolation part. When this criterion is selected in a very specific manner, we recover the classical GLRT of the literature, while we obtain interesting novel tests with alternative criteria. Our mathematical derivations are surprisingly simple considering that they solve a problem that has been open for more than half a century.

SPEAKER: Dr. Gregory Abowd

ABSTRACT:

Research and education in health systems at Georgia Tech has a long history, but there has been a definite surge in interest over the past five years. The Health Systems Institute was established in 2006 as a means of coalescing a variety of different research perspectives between Georgia Tech and Emory. In this talk, I will give an overview of the mission of HSI and describe the variety of projects that various faculty and students are engaged in. Featured in this talk will be several major efforts in partnership with Children's Healthcare of Atlanta.

BIO
Gregory Abowd is a Distinguished Professor in the School of Interactive Computing and the Interim Director of the Health Systems Institute. In his 14-year career at Georgia Tech, he has worked on applications of ubiquitous computing technology in real environments, with landmark projects Classroom 2000 and the Aware Home. Through his involvement with the Aware Home, he and others in GVU and the School of Interactive Computing have looked at applications that impact important health questions around chronic care management. In his new role with HSI, he is leading a much broader research agenda to see how to impact all aspects of healthcare delivery.

SPEAKER: Dr. Eldad Haber

ABSTRACT:

While experimental design for well-posed inverse linear problems has been well studied, covering a vast range of well-established design criteria and optimization algorithms, its ill-posed counterpart is a rather new topic. The ill-posed nature of the problem entails the incorporation of regularization techniques. The consequent non-stochastic error introduced by regularization, needs to be taken into account when choosing an experimental design. We discuss different ways to define an optimal design that controls both an average total error of regularized estimates and a measure of the total cost of the design. We also introduce a numerical framework that efficiently implements such designs and natively allows for the solution of large-scale problems. To illustrate the possible applications of the methodology, we consider a borehole tomography example and a two-dimensional function recovery problem.

Contact: ISyE DOS Optimization Seminars (http://www2.isye.gatech.edu/dos/)

SPEAKER: Dr. Prasad Tetali

ABSTRACT:

The notion of a correlation decay, originating in statistical physics, has recently played an important role in yielding deterministic approximation algorithms for various counting problems. I will try to illustrate this technique with two examples: counting matchings in bounded degree graphs, and counting independent sets in certain subclasses of claw-free graphs.

SPEAKER: Dr. Mark Braunstein

ABSTRACT:

Georgia Tech and Emory are discussing the formation of a new School of Biomedical Informatics. This talk will provide some of the background for the field and will review its principle domains; it will explain why one of the key domains

SPEAKER: Dr. Zimin Lu
Head of Quantitative Research
Financial Products, BP Energy Company

ABSTRACT:

There are three value buckets for a storage asset holder: a) Winter-Summer spreads due to demand seasonality; b)Monthly injection/withdrawal optionalities; c) Intra-month inventory optionalities. The valuation problem involves both derivative pricing and financial operations. I will discuss an approach based on the constrained portfolio optimization of time spread options.

Bio:
Dr. Zimin Lu is currently the head of Quantitative Research for North America Gas and Power, BP Energy Company in Houston. His energy quantitative finance experience includes Vice President at Deutsche Bank in New York; Vice President of Research & Analysis at Suez North America; Director at Enron Research and Quantitative Analyst at TXU Energy.

He earned his Ph.D. in Physics from Drexel University and B.S. from Nanjing University. He worked for University of Oregon and Princeton University during the transition from academia to energy industry.

SPEAKER: Alejandro MacCawley

ABSTRACT:

The acquisition process for raw materials in the flower industry involves a large number of technical, biological and economic variables. We develop a Bulb Acquisition Planning (BAP) mixed integer optimization model (MIP). The model incorporates aspects such as substitutability, precocity, risk diversification, seasonality, and inventory among others. The model was implemented using enterprise data from production planning executed in the past season. The results of the BAP model and the executed plan are compared in this talk with respect to cost, container programming and inventory maintenance.

SPEAKER: Michael Hamada
Statistical Sciences Group
Los Alamos National Lab

ABSTRACT:

The practice of statistics involves analyzing data and planning data collection to answer scientific/engineering questions. Issues often arise which must be dealt with and can lead to new results, i.e.,research. In analyzing data, these issues can sometimes be addressed in the statistical models that are developed. Simulation can be helpful in evaluating a proposed procedure. Moreover, simulation coupled with an optimization algorithm can be used to plan a data collection scheme. I will illustrate these points with some simple experiments motivated by my Dad and two daughters. It turns out that conducting a simple experiment is not so simple and almost immediately one is faced with the practice of statistics.

SPEAKER: Dr. Hein van der Holst

ABSTRACT:

Each graph can be embedded in 3-space. The problem becomes more interesting if we put restrictions on the type of embedding. For example, a linkless embedding of a graph is one where each pair of vertex-disjoint circuits has linking number equal to zero. The class of all graphs that have a linkless embedding is closed under taking minors. Robertson, Seymour, and Thomas gave the forbidden minors for this class of graphs. Open remained how to find a linkless embedding in polynomial time. In the talk we start with discussing an algorithm to find a linkless embedding.

Instead of embedding the graph in 3-space, we could also consider mapping properties of certain superstructures of the graph in 3-space, and, indeed, if this superstructure has not the right mapping properties in 3-space, see whether it has the right one in 4-space, etc. We introduced for a graph G a new graph parameter (G), which is defined as the smallest d such that superstructures of G have a zero intersection mapping in d-space. The nicest property of this graph parameter is its independence of the superstructure and thus depends on the graph only. For d = 2 and d = 3, (G) d if and only if G is outerplanar and planar, respectively. The graphs G with (G) 4 are exactly those that have a linkless embedding. In the second part of the talk we will discuss this new graph parameter. (This part is joint work with R. Pendavingh.)

SPEAKER: Guanghui Lan

ABSTRACT:

We consider an important class of convex programming problems whose objective functions, given by the summation of a smooth and non-smooth component, are contaminated by stochastic noise. Although a valid lower bound on the rate of convergence for solving these problems is known from the classic complexity theory of convex programming due to Nemirovski and Yudin, the optimization algorithms that can achieve this lower bound have not been discovered. In this talk, we show that the robust stochastic approximation method exhibits the best-known rate of convergence for solving these problems. The main goal is to present the first theoretically optimal method for solving this class of problems that achieves the aforementioned lower bound on the rate of convergence and demonstrate its significant advantages over existing algorithms.

SPEAKER: Dr. Abhyuday Mandal
Department of Statistics
University of Georgia

ABSTRACT:

Identifying promising compounds from a vast collection of feasible compounds is an important and yet challenging problem in pharmaceutical industry. An efficient solution to this problem will help reduce the expenditure at the early stages of drug discovery. In an attempt to solve this problem, Mandal, Wu and Johnson (2006) proposed the SELC algorithm. Although powerful, it fails to extract substantial information from the data to guide the search efficiently as this methodology is not based on any statistical modeling. The proposed approach uses Gaussian Process modeling to improve upon SELC, and hence named G-SELC. The performance of the proposed methodology is illustrated using four and five dimensional test functions. Finally, we implement the new algorithm on a real pharmaceutical data set for finding a group of chemical compounds with optimal properties.

(Joint research with C.F. Jeff Wu and Pritam Ranjan)

SPEAKERS: Pinar Keskinocak and Ellen Do

ABSTRACT:

In the U.S. alone, nosocomial (hospital-acquired)infections occur at the rate of approximately two million per year, with one hundred thousand of these leading to death. The suffering, loss of productivity, expenditure of scarce health-care resources, coupled with the rising resistance of many pathogens, make this issue urgent and important. Since it is difficult or expensive to observe colonization and ascertain whether a patient was infected prior to, or during hospitalization, and the process by which pathogens spread depends on the behavior of patients, health-care workers, visitors and the community at large, the best strategy to reduce nosocomial infection rates is unknown. Two possible approaches to reduction include: screening for isolation and improved hygiene approaches. The costs involved are not yet clear, and in particular, the non-healthcare productivity costs need to be included in any cost-benefit analysis of policies.

Pinar Keskinocak will discuss her research on Evaluating the Efficacy and Efficiency of Competing Approaches to Reducing Nosocomial Infection Rates, using simulation to model how different strategies may impact rates of nosocomial infections, their costs in the long and short term, as well as the overall interaction between hospitals and the community. Improved adherence to hand hygiene can significantly reduce the transmission of multi-drug resistant organisms and prevent many healthcare associated infections. Despite widespread acceptance of the importance of hand hygiene and the existence of CDC/HICPAC guidelines recommending hand hygiene before and after patient contact, compliance with hand hygiene is poor in many healthcare settings. Successful programs to increase hand hygiene compliance are multifaceted and include a component of compliance monitoring. Electronic monitoring and reminding systems show considerable promise in providing significant and lasting improvements in hand hygiene.

Ellen Yi-Luen Do will discuss the project of Nosocomial Control using RFID technology with Evidence-Based Design for patient rooms, as well as the Technological Interventions and Measurement for Hand Hygiene Adherence project to create a next generation monitoring system that will allow more accurate automatic data measurement and recording, to be implemented at Emory Crawford Long Hospital pending on CDC funding.

SPEAKER: Tony Cai
Department of Statistics
The Wharton School
University of Pennsylvania

ABSTRACT:

Much of the nonparametric regression theory is focused on the case of additive Gaussian noise. In such a setting many smoothing techniques including wavelet thresholding methods have been developed and shown to be highly adaptive.

In this talk we consider robust nonparametric regression, where the noise distribution is unknown and possibly heavy-tailed, and generalized nonparametric regression in exponential families which include, for example, Poisson regression, binomial regression, and Gamma regression. We take a unified approach of using a transformation to convert each of these problems into a standard homoskedastic Gaussian regression problem. Then in principle any good nonparametric Gaussian regression procedurecan be applied to the transformed data. We use a wavelet block thresholding procedure to illustrate our method and show that the resulting estimators are adaptively rate-optimal over a range of Besov Spaces. The procedure is easily implementable. A key technical step is the development of a quantile coupling theorem that is used to connect our problem with a more familiar Gaussian setting.

SPEAKER: Dr. Nagi Gebraeel

ABSTRACT:

This presentation considers a fleet of identical units (a fleet of aircrafts, trucks, etc.) where each unit in the fleet consists of several key components that are critical to the operation of a unit. We assume that each of these components degrades gracefully over time and its degradation process can be modeled using parameterized stochastic degradation models. The presentation discusses a novel approach that identifies the top-k, the k units with the largest residual life (most reliable), using sensor- based prognostic information associated with the critical components of each unit. We propose a Prognostics-Based Ranking Algorithm (PBR-Algorithm) that combines prognostics-based stochastic degradation models with computer science database ranking algorithms. The stochastic degradation models are used to compute residual life distributions (RLDs) of partially degraded critical components using condition monitoring data, whereas the ranking algorithm orders the units using the critical component as a ranking predicate and the Median of the RLD as a scoring criterion. The stochastic models used in this work utilize in-situ
degradation signals to continuously update the RLDs, in real-time. This creates a set of dynamically evolving RLDs. The non-parametric nature of the RLDs coupled with the need to perform real-time computations pose significant challenges in identifying the subset of k units with the largest remaining useful life. In this presentation, we will show that given the time stamp and amplitude/level of a component's degradation signal, it is possible to establish stochastic dominance among the RLDs of all the units in the fleet (for a given component type). We also
demonstrate the integration of the stochastic ordering results with advanced ranking algorithm-used for ranking of database queries in computerized search engines-for identifying the top-k units.

SPEAKER: Antonio Carbajal

ABSTRACT:

LTL carriers can attain cost savings by correctly identifying the types of trailer equipment to use on different linehaul service lanes. To date,research has focused only on short haul equipment mix problems in routing and scheduling problems. Alternatively, we focus on linehaul equipment mix. Given a load plan, we develop models to determine which direct trailer moves should be loaded with pups (28

SPEAKER: Dr. Jose Correa

ABSTRACT:

We study a game that models a market in which heterogeneous producers make pricing decisions in a first stage, followed by consumers that select a producer selling at lowest price. Producers submit a price function to the market, which maps a production level to a price. Solutions of this type of models are normally referred to as supply function equilibria, and the most common application is in electricity markets. In our model, producers have increasing marginal production costs equal to a linear function, and adopt price functions that are proportional to their production costs. Therefore, the decision in the first phase of the game is the markup producers apply to the production cost. In a second phase of the game, consumers learn the producers' price functions, which leads to an allocation of consumers to producers.

We prove that an equilibrium exists if and only if there are at least three producers, and if an equilibrium exists, it is unique. We also establish that for highly competitive markets, the equilibrium assignment is nearly efficient, where an efficient assignment is one that minimizes the total production cost. Furthermore, we establish an almost tight bound on the worstcase inefficiency of an equilibrium. In particular, the bound states that when there are two equally efficient producers and possibly other less efficient ones, the production cost under an equilibrium is at most 50% worse than the optimal one, and the worst-case gap between the two assignments decreases rapidly as competition increases. For instance, for three similarly-efficient producers plus perhaps other less efficient ones, the inefficiency is below 6.2%.

This is joint work with Nicolas Stier-Moses.

SPEAKER: Carlos F. Daganzo
Director, Center for Future Urban Transportation
Institute of Transportation Studies
University of California, Berkeley

ABSTRACT:

It is shown that under some conditions some properties of a road link -- such as its capacity and its fundamental diagram -- scale up when several links are joined together to form a regular structure (e.g., a ring). Of particular interest is the existence of both, a macroscopic fundamental diagram (MFD) relating a network

SPEAKER: Dr. Stephan Held (University of Bonn)

ABSTRACT:

A central characteristic of a computer chip is the speed at which it processes data, determined by the time it takes electrical signals to travel through the chip. A major challenge in the design of a chip is to achieve timing closure, that is to find a physical realization fulfilling the speed specifications.

We give an overview over the major tasks for optimizing the performance of computer chips and present several new algorithms. For the topology generation of repeater trees, we introduce a variant of the Steiner tree problem and present fast algorithm that balances efficiently between the resource consumption and performance.

Another indispensable task is gate sizing, a discrete optimization problem with nonlinear or PDE constraints, for which a fast heuristic is introduced. The effectiveness in practice is demonstrated by comparing with newly developed lower bounds for the achievable delay.

We conclude with a variant of the time-cost tradeoff problem from project management. In contrast to the usual formulation cycles are allowed. We present a new method to compute the time-cost tradeoff curve in such instances using combinatorial algorithms. Several problems in chip design can be modeled as time-cost tradeoff problems, e.g. threshold voltage optimization of plane assignment.

SPEAKER: John Trestrail

ABSTRACT:

Each year the US Department of Agriculture (USDA) and the US Agency forInternational Development (USAID) spend approximately $2 billion on US agriculture products and ocean transportation to combat world hunger. The US Government-sponsored international food aid programs is a unique alignment between US Government agencies, NGO like CARE and Save the Children, the World Food Programme, US Agri-business, and the international shipping and supply chain business community.

Atlanta-based Next Wave Agency, LLC plays a small role in the process by working on the behalf of some of the ocean carriers that transport the bagged food from US ports to foreigncountries for distribution. Next Wave serves their client in the formal role of a ship broker with duties that include advising clients of the market conditions, explaining government regulations, and submitting freight offers in response to invitations to bid to the US government.

The US Department of Agriculture procures the food and transportation services under the mandate of minimizing the

SPEAKER: Juan Pablo Vielma

ABSTRACT:

We review several Mixed Integer Programming formulations for separable continuous piecewise linear cost functions and their extensions to the non-separable and/or discontinuous cases. We compare both existing and new formulations by their theoretical properties and by their performance in computational experiments. We present results for different classes of transportation problems with separable/non-separable, continuous/lower-semicontinuous non-convex piecewise linear cost functions.

This is joint work with Prof. Shabbir Ahmed and Prof. George Nemhauser.

SPEAKER: Pitu Mirchandani

ABSTRACT:

Real-time Decision Systems include systems such as real-time traffic signal control systems, real-time transport scheduling systems, real-time freight dispatch systems, real-time inventory control, etc. Given the vast amount of streaming data that is available through sensors for operating such systems, one can make better real-time decisions if such data is used appropriately. "Real-time" necessarily implies that it is some sort of feedback control, where new decisions are implemented based on "feedback" from sensors that measure the "state of the system". As such, the attributes of the algorithms embedded in the feedback influence how well the system responds to scheduled and unscheduled external events.

This talk will first present a framework for the design of a real-time adaptive systems and some current work in traffic adaptive signal control systems require that various traffic data be collected in order to develop a traffic model within which the desired performance optimization can be completed. These data include link travel times, turn proportions and queue discharge rates. A brief discussion of a system developed by the presenter, referred to as RHODES, will be also discussed.

The second part of the talk will discuss the application of the framework for the design of a wide-area adaptive traffic management for evacuations

SPEAKER: Dr. Tim Randolph

ABSTRACT:

In a 'functional linear model' the aim is to estimate the linear relationship between a scaler response, y, and a random L2 function, x. This formally fits into the classical regression model, y = X*b + e, but the estimation of b is inherently ill-posed. A common method used to address this problem regularizes the solution by penalizing L*b, where L is a linear operator. This constrains b (which is nominally any L2 function) by bounding the norm of L*b. Special cases include Marquardt's generalized regression, Hoerl's ridge regression (L=I) and

Tikhonov-Philips' regularization (L=D, a differential operator). This talk considers (linear) penalized estimates in general and their expression via generalized singular vectors of the pair (X,L). This provides perspective on the functional structure of an estimate and on criteria for choosing the penalty operator. Corresponding explicit expressions for bias and variance provide insight regarding the trade-off between the properties of L versus X.

SPEAKER: Dr. Michael L. Best

ABSTRACT:

We have been investigating the role of rich new information and communication environments in efforts to restore and reconcile a people or nation destroyed by years of civil strife and conflict. Through an integrated and holistic approach, we have collaboratively designed, implemented, deployed and evaluated interactive environments that support dialogue, deliberation and memory for the resolution of civil conflict. The systems have included rich online web presences, offline interactive multi-modal computer environments, and text and voice based mobile phone systems.

Our current work is in collaboration with the Truth and Reconciliation Commission (TRC) of Liberia, though we are exploring new projects, for example with the traditional Cultural Institution of the Acholi people, Northern Uganda. I will report on this ongoing work and try to extend our findings to current and emerging conflicts.

Bio: Dr. Michael L. Best is assistant professor at the Sam Nunn School of International Affairs and the School of Interactive Computing at Georgia Institute of Technology where he is a research director with their GVU Center. He is also a Fellow of the Berkman Center for Internet & Society at Harvard University. Professor Best is co-founder and Editor-in-Chief of Information Technologies and International Development published by the MIT Press. He is a frequent consultant to the World Bank, ITU, and USAID. He holds a Ph.D. from MIT and has served as Director of Media Lab Asia in India and head of the eDevelopment group at the MIT Media Lab

SPEAKER: Dr. Guy Lebanon

ABSTRACT:

Statistical models on full and partial rankings of n items are often of limited practical use for large n due to computational consideration. We explore the use of non-parametric models for partially ranked data and derive computationally efficient procedures for their use for large n. The derivations are largely possible through combinatorial and algebraic manipulations based on the lattice of partial rankings. A bias-variance analysis and an experimental study demonstrate the applicability of the
proposed method.

Bio:

Guy Lebanon is an assistant professor of computing at the Georgia Institute of Technology. His main research area is statistical modeling and visualization of high dimensional discrete data such as text documents and partially ranked data. Additional research interests include privacy preservation in databases and social networks and the
use of non-Euclidean geometry in machine learning.

Prior to his current appoitnment at Georgia Tech, Dr. Lebanon was an assistant professor of statistics and electrical and computer engineering at Purdue University. He recieved his PhD in 2005 from Carnegie Mellon University and BA, MS degrees from Technion - Israel Institute of Technology, all in computer science.

Prof. Lebanon received the NSF CAREER Award, Purdue's Teaching for Tomorrow Award, the 2004 LTI SRS Best Presentation Award and is a Siebel scholar.

SPEAKER: Dr. Harold W. Kuhn (Professor Emeritus, Princeton University)

ABSTRACT:

Fifty five years ago, two results of the Hungarian mathematicians, Koenig and Egervary, were combined using the duality theory of linear programming to construct the Hungarian Method for the Assignment Problem. In a recent reexamination of the geometric interpretation of the algorithm (proposed by Schmid in 1978) as a steepest descent method, several variations on the algorithm have been uncovered, which seem to deserve further study.

The lecture will be self-contained, assuming little beyond the duality theory of linear programming. The Hungarian Method will be explained at an elementary level and will be illustrated by several examples.

We shall conclude with account of a posthumous paper of Jacobi containing an algorithm developed by him prior to 1851 that is essentially identical to the Hungarian Method, thus anticipating the results of Koenig (1931), Egervary (1931), and Kuhn (1955) by many decades.

Speaker: Dr. Shannon Barker

Recent exciting advances in genetic technologies are driving the emergence of modern medical approaches, specifically diagnostics and gene transfer, for improved human health. Our increasing understanding of the role of DNA in health and disease, through human genome sequencing and breakthroughs in gene manipulation, delivery, and analysis technologies, is leading to improved diagnostics and more effective and personalized treatments for patients. While highly promising and potentially cost-saving, these technologies present many technical and policy challenges, and come with a multitude of complicated issues regarding translation into actual health care.

For example, the regulatory mechanisms of genetic tests in the U.S. have significant gaps in comprehensive oversight that exists within a wide range of regulatory agencies. Little evidence-based research has been performed for determining the affects of these gaps on test development, implementation, and ultimately, patient care. The development of appropriate oversight and regulation to address technological, societal, and ethical concerns is crucial for both the innovation of the genetic testing field and the protection of patient safety.

Another genetic technology rapidly approaching reality is human germline genetic modification (HGGM). HGGM refers to permanently altering one

SPEAKER: Dr. Kartik Sivaramakrishnan (Researh Div., Axioma,Inc.)

ABSTRACT:

Semidefinite programming (SDP) is referred to as "linear programming for the 21st century" and has a variety of applications in science and engineering. Primal-dual interior point algorithms are currently the most popular techniques for solving large scale SDPs. However, these algorithms are fairly limited in the size of SDPs that they can solve in practice.

We present a two stage decomposition algorithm to solve large scale structured semidefinite programs (SDPs). In the first stage, we exploit the sparsity and/or the symmetry in an underlying SDP in order to process it into an equivalent problem having a "block-angular" structure. We will illustrate the procedure with simple examples in the talk. In the 2nd stage, we solve the resulting "block-angular" SDP in an iterative fashion between a master problem and "decomposed" and "distributed" subproblems in a parallel computing environment. We will give the details of our decomposition algorithm and also highlight some of our enhancements that improve the convergence of the algorithm. We will also report our computational experiences with the algorithm on the distributed "Henry2" cluster at NC State University. Finally, we compare our algorithm with the OpenMP version of CSDP that is available via the COIN-OR interface.

SPEAKER: Dipl.-Ing. Detlef Spee

ABSTRACT:

Order picking is a crucial process for the efficiency of warehouses. Already the largest component of labor cost in the warehouse, order picking systems are being stressed because customers are ordering more frequently and the orders are becoming smaller. For this reason efficiency in order picking is becoming more critical. Quality, efficiency, the right throughput time for order picking are strongly influenced by the selection of the order picking system. In this talk, examples for selection criteria will be given and application examples will be presented that show how order picking systems have been designed by the combination of well known material handling components. The talk also will address the interactions between order picking requirements, order picking concepts the current direction of order picking automatization.

About the speaker: Detlef Spee received the Dipl.-Ing. from the University of Dortmund in 1985, where he then worked as a scientist in conveyor technology for a year. Subsequently, he joined the Fraunhofer IML, where he has been a Project Engineer, Head of the Handling Technology and Order Picking Systems Department, and since 1997, Head of the Material Handling Engineering Department. He is a frequent lecturer to business and industry on material handling system design and warehouse management systems. Detlef has been a key partner to Georgia Tech in the student exchange program with the University of Dortmund, and an active participant in the International Material Handling Colloquium.

About the Fraunhofer IML: The Fraunhofer IML acts as a logistic consultant and applied research organization for companies of all lines of industry and sizes. Our teams help clients to tackle new tasks and requirements. Together with clients we optimize the internal and external logistics and develop software and hardware solutions. Fraunhofer IML favours company-specific and tailormade solutions and supports its customers from the planning up to the realisation. Various products, like palletizing robots, warehouse systems or simulation software, were first developed in Dortmund and are now used worldwide.

The institute is structured into three main sections:
Material Flow Systems
Enterprise Logistics and
Logistics, Traffic and Environment

SPEAKER: Professor Achal Bassamboo (Northwestern Univ.)

ABSTRACT:

We study a capacity sizing problem in service systems with uncertain arrival rates; telephone call centers are canonical examples of such systems. The objective is to choose a staffing level that minimizes the sum of personnel costs and abandonment/waiting time costs. We formulate a simple fluid analogue, which is in essence a newsvendor problem, and demonstrate that the solution it prescribes performs remarkably well. In particular, the gap between the performance of the optimal staffing level and that of our proposed prescription is independent of the "size" of the system, i.e., it remains bounded as the system size (demand volume) increases. This stands in contrast to the more conventional theory that applies when arrival rates are known, and commonly used rules-of-thumb predicated on it. Specifically, in that setting the difference between the optimal performance and that of the fluid solution diverges at a rate proportional to square-root of the size of the system. One manifestation of this is the celebrated square root safety staffing principle that dates back to work of Erlang, which augments solutions of the deterministic analysis with additional servers of order square root the volume of demand. In our work, we establish that this type of prescription is needed only when arrival rates are suitably "predictable."

SPEAKER: Aly Megahed

ABSTRACT:

Much research has been conducted in the area of supply chain network design. Problems and challenges addressed include the location of facilities, allocation of customer demand to different facilities, multi-commodity planning, inventory accumulation, and production decisions. However, the majority of research efforts address these problems separately, or consider only two or three simultaneously.

In this presentation, we present a mixed integer linear programming (MILP) model that addresses the above mentioned challenges in a single model. The result is a capacitated three echelon multi-commodity dynamic production- distribution model with inventory accumulation onsideration (CTMDPI). The study was motivated by and implemented for gsf Egypt, a multinational industrial company that supplies McDonalds and many fast food restaurants in Egypt with most of their requirements.

SPEAKER: Dr. Anureet Saxena (Research division, Axioma, Inc.)

ABSTRACT:

Convex Relaxations of Non-Convex Mixed Integer Quadratically Constrained Programs: Extended Formulations
http://www.optimization-online.org/DB_HTML/2008/08/2068.html

This paper addresses the problem of generating strong convex relaxations of Mixed Integer Quadratically Constrained Programming (MIQCP) problems. MIQCP problems are very difficult because they combine two kinds of non-convexities: integer variables and non-convex quadratic constraints. To produce strong relaxations of MIQCP problems, we use techniques from disjunctive programming and the lift-and-project methodology. In particular, we propose new methods for generating valid inequalities by using the equation Y = xx^T. We use the concave constraint 0 succcurlyeq Y - xx^T to derive disjunctions of two types. The first ones are directly derived from the eigenvectors of the matrix Y - xx^T with positive eigenvalues, the second type of disjunctions are obtained by combining several eigenvectors in order to minimize the "width" of the disjunction. We also use the convex SDP constraint Y - xx^T succcurlyeq 0 to derive convex quadratic cuts, and we combine both approaches in a cutting plane algorithm. We present computational results to illustrate our findings.

Convex Relaxations of Non-Convex Mixed Integer Quadratically Constrained Programs
http://www.optimization-online.org/DB_HTML/2008/11/2145.html

A common way to produce a convex relaxation of a Mixed Integer Quadratically Constrained Program (MIQCP) is to lift the problem into a higher dimensional space by introducing variables Y_{ij} to represent each of the products x_i x_j of variables appearing in a quadratic form. One advantage of such extended relaxations is that they can be efficiently strengthened by using the (convex) SDP constraint Y - x x^T succeq 0 and disjunctive programming. On the other hand, their main drawback is their huge size, even for problems of moderate size. In this paper, we study methods to build low-dimensional relaxations of MIQCP that capture the strength of the extended formulations. To do so, we use projection techniques pioneered in the context of the lift-and-project methodology. We show how the extended formulation can be algorithmically projected to the original space by solving linear programs. Furthermore, we extend the technique to project the SDP relaxation by solving SDPs. In the case of an MIQCP with a single quadratic constraint, we propose a subgradient-based heuristic to efficiently solve these SDPs. We also propose a new "eigen reformulation" for MIQCP, and a cut generation technique to strengthen this reformulation using polarity. We present extensive computational results to illustrate the efficiency of the proposed techniques. Our computational results have two highlights. First, on the GLOBALLib instances, we are able to generate relaxations that are almost as strong as those proposed in our companion paper even though our computing times are about 100 times smaller, on average. Second, on the box QP instances, the strengthened relaxations generated by our code are almost as strong as the well-studied SDP+RLT relaxations and can be solved in less than 2 sec even for larger instances with 100 variables; the SDP+RLT relaxations of the same set of instances can take up to a couple of hours to solve using a state-of-the-art SDP solver.

This is joint work with Pierre Bonami and Jon Lee

Chris Schlegel received his BA and MA in economics at Simon Fraser University, British Columbia. Following his graduate work he was Research Economist at The Fraser Institute, and from 2001 served as expert witness for the cost of capital and market analysis for the staff of the New Hampshire Public Utilities Commission in Concord, NH.

Chris subsequently worked as Senior Analyst and later Manager in the Risk Control group on the trading floor of Southern Company, where he developed power and gas trading risk metrics and portfolio valuation tools. He is presently Manager of Risk Analysis Services, a group at Southern Company headquarters in Atlanta focused on quantitative risk management.

Chris is President of the U.S. Association for Energy Economics in Atlanta, and was recently named Associate at the James C. Bonbright Center for Public Utilities at the University of Georgia. He has been invited as speaker to numerous professional conferences on quantitative risk modeling and utility economics, and has published in peer reviewed journals including Energy Risk and the European Journal of Political Economy.

His professional interests include energy economics, quantitative and behavioral finance, and enterprise risk management.

Ally Hung is a Staff Financial Analyst with Enterprise Risk Management. Ally graduated from the University of Tennessee, Knoxville with a Master in Economics and a Bachelor in Finance.

She joined the Risk Analysis Service (RAS) group in March 2008 from SCS Forecasting. Her current areas of responsibilities include: Modeling and analyzing risks (commodity price and interest rate), identify and assess risks, communicate potential risk mitigation opportunities using financial instruments. Before joining the RAS, Ally was a forecaster at SCS Forecasting with focus on the development of the Energy and Peak Demand forecast using structural Long Term End-Use models.

Prior to joining Southern, she was a Stock Trader in Hong Kong and holds a trader license from the Hong Kong Stock Exchange. Ally also serves as the Logistic Chair for the APA.

ABOUT THE EVENT:

Students representing five disciplines (architecture, health systems, human-computer interaction, computer science, and systems engineering) will showcase their designs. You will have the opportunity to view the prototypes and talk with the instructors and student teams about the projects.

----------------------
THE PROJECTS
----------------------
- Aftercare Support System Design
- Care on Rails: Extendable Adaptable Sliders
- FamLink
- Balek Lighting: Integrated Lighting System
- Mpower: The New Patient Lifeline
- Open Space
- Personal Waiting Room Assistant
- Triar: The Triage Chair

----------------------
THE INSTRUCTORS
----------------------
- Jeremy Ackerman, MD PhD, Assistant Professor of Emergency Medicine, Emory University
- David Cowan, Program Director and Executive in Residence at the Health Systems Institute
- Ellen Yi-Luen Do, PhD, Associate Professor, Joint Appointments in College of Architecture & College of Computing
- Marilyn Margolis, RN, MN, Director of Operations, Emory University Hospital
- Marvina Williams, RN, Perkins +Will Architects
- Craig Zimring, PhD, Professor of Architecture and Psychology

SPEAKER: Timothy A. Thornton, Ph.D.
UC President's Postdoctoral Fellow
Department of Statistics
University of California, Berkeley

ABSTACT:

Genetic association testing has proven to be a valuable tool for the mapping of complex traits. Technological advances have made it feasible to perform case-control association studies on a genome-wide basis. Some of the characteristics of the data include missing information, and the need to analyze hundreds of thousands or millions of genetic markers in a single study, which puts a premium on computational speed of the methods. The observations in these studies can have several sources of dependence, including population structure and relatedness among the sampled individuals, where some of this structure may be unknown. Neglecting such structure in the data can lead to seriously spurious associations. We describe a new approach to this problem.

SPEAKER: Dr. Francois Margot (CMU and Axioma,Inc.)

ABSTRACT:

Many industrial problems can be naturally formulated using Mixed Integer Nonlinear Programming (MINLP). Motivated by the demand for Open-Source solvers for real-world MINLP problems, we have developed a spatial Branch-and-Bound software package named couenne (Convex Over- and Under-ENvelopes for Nonlinear Estimation). In this talk, we present the structure of couenne and discuss in detail our work on two of its components: bounds tightening and branching strategies. We then present experimental results on a set of MINLP instances including some industrial applications. We also compare the performance of couenne with a state-of-the-art solver for nonconvex MINLPs.

Joint work with P. Belotti (Lehigh); L. Biegler, G. Cornuejols, I. Grossman (CMU); P. Bonami (LIF, Marseille); J. Lee, A. Waechter (IBM); L. Liberti (LIX, Paris).

SPEAKER: Professor Johne Parker

ABSTRACT:

Current advances in science and engineering increasingly require a multi-faceted approach best addressed by the collaboration of researchers from various technical disciplines. Interdisciplinary research is no longer a novel concept; collaborations across complementary technical disciplines have become common and will continue to be increasingly important. However, many of our
current technical challenges are also coupled with complex legal, societal and policy issues; these problems will best be solved by engineers who are also conversant in the language of policy (and may be in collaboration with scholars who have a background in policy or law).

There are few better places to truly learn about public policy than our nation