Many simple, deterministic systems may exhibit extremely complicated behavior (called chaotic behavior). The variability induced by the chaotic behavior of a system can be so large that the output of the system is effectively random. This variability may severely deteriorate the performance of a system in a manufacturing environment. By studying the dynamical behavior of the system this chaotic behavior may be avoided with a proper system configuration. We study a generalized model of bucket brigades in which workers walk back with finite velocities and are allowed to pass each other. Based on this model, we have formally proved that the dynamics of a purely deterministic bucket brigade assembly line can be chaotic and may induce large variability if it is improperly configured.
We present a sufficient condition to avoid this chaotic behavior and to make the bucket brigade self-balanced.

The bucket brigade protocol can be adapted to assembly "trees". In a network of subassembly lines in which subcomponents are produced and are then successively combined to make the final product, balance becomes more difficult to achieve as it requires that all subassembly lines be synchronized to produce at the same rate. We show how to adapt the bucket brigade protocol of work-sharing so that balance emerges spontaneously.

We have also generalized the idea of bucket brigades to flow lines with multiple job sources. This work is motivated by the order-picking operation in a major distribution center in which customer orders arrive arbitrarily in time and are introduced into the system at multiple points along an aisle. We compare the performance of an adapted bucket brigade protocol with the order-picking policy currently adopted by the distribution center.

First, we propose enhancements to an existing outer-approximation algorithm -- called the Polyblock algorithm -- for monotonic optimization problems. The enhancements are shown to significantly improve the computational performance of the algorithm while retaining the convergence properties. Next, we develop a generic branch-and-bound algorithm for monotonic optimization problems. A computational study is carried out for comparing the performance of the Polyblock algorithm and variants of the proposed branch-and-bound scheme on a family of separable polynomial programming problems. Finally, we study an important class of monotonic optimization problems -- probabilistically constrained linear programs. We develop a branch and bound algorithm that searches for a global solution to the problem. The basic algorithm is enhanced by domain reduction and cutting plane strategies to reduce the size of the partitions and hence tighten bounds. The proposed branch-reduce-cut algorithm exploits the monotonicity properties inherent in the problem, and requires solving linear programming subproblems. We provide convergence proofs for the algorithm. Some illustrative numerical results involving problems with discrete distributions are presented.

In this talk, I will present optimization models for this type of problems. I will discuss formulating a proper objective function and problem structure that facilitates the search for the optimal solution. I will also discuss the use of optimization models to construct customer admission policies.

Speaker Bio:

Qiong Wang has a Ph.D in Engineering and Public Policy, a master degree in Electrical and Computer Engineering, both from Carnegie-Mellon University.
He is currently a Member of Technical Staff with Industrial Mathematical and Business Analysis Group in Bell Labs' Mathematical Science Center. He research area is pricing, capacity planning, and revenue management with application to communication networks. His modeling work includes multi-stage network optimization, bandwidth provisioning in the presence of demand uncertainty, joint planning of pricing, capacity deployment, and network engineering, as well as asymptotical analysis of revenue management policy. Recently he has been working on coordination mechanisms for managing technology innovation.

Our cost structure incorporates partial reversibility of investment by differentiating the buying price and the salvage value of the capacity. This partial reversibility makes the value function non-differentiable and divides the solution space into BUY, KEEP, and SELL regions. In addition, with the non-differentiable value function, it is hard to obtain an analytical solution in general. By identifying some properties of the solution structure, we can perform a series of sensitivity analyses of optimal investment decisions with other market parameters. Under a typical condition, the cost depreciation is steeper than the movement of the economic condition, which shows the following: Capital investment can be analytically expressed. The multiple-time investment decision problem is reduced to a one-time investment decision. Increases are dependent on the economic condition of the current period and not on expected future economic conditions.

We study this problem in both the monopolistic and oligopolistic markets. In particular, we investigate investment decision of firms in duopoly markets, assuming that firms follow Cournot competition behavior. We attempt to prove the existence and the uniqueness of the Cournot equilibrium point in a duopoly market. In

addition, we show how competition between firms affects the market properties through total market capacity, market price, consumer surplus, expected time to a certain price reduction, and the expected time to the first investment decision. Finally, we perform a linear regression analysis using market data and attempt to validate our model by showing the relationship between capacity expansion and economic indicators.

Generally, the derivation of the inventory policies requires the knowledge of the specific demand distribution. Unfortunately, in several cases the demand is not observable in a direct way. The second problem is motivated by a practical application where only partial demand information is observable. Towards this end we derive estimators of the first two moments of the (daily) demand by means of the renewal theoretical concepts. We also propose a regression-based approximation to improve the quality of the estimators. A series of numerical studies are carried out to evaluate the accuracy and precision of the estimators and to investigate the impact of the estimation on the optimality of the inventory policies.

The last part of this dissertation studies a periodic-review inventory system with regular and emergency orders. Emergency orders, characterized by shorter lead-time, higher ordering cost and higher setup cost, are placed when the inventory level becomes critically low. Based on our assumptions, we formulate a dynamic programming model and prove the optimality of $(s, S)$ type polices for both emergency and regular orders. We also derive analytic properties of the optimal policies. We gain some managerial insights into the optimal policies and cost performance from numerical studies.

TI:GER is a multi-disciplinary, engaged learning program intended for students interested in understanding the innovation processes that are used in commercializing technologies. The program is a unique educational collaboration between Georgia Tech and Emory University which prepares students for the challenges of commercializing new technology and delivering innovative products to the marketplace. It is a learning experience where engineering and science PhD students are matched with Georgia Tech's College of Management MBA students and Emory University law students. These teams work together to develop an understanding of the technical, market, and legal challenges of commercializing the PhD student's primary research. A primary goal of TI:GER is to help produce fundamental research with both scientific merit and market relevance.

Abstract:
We analyze the structure of a truckload transportation contract that encourages information and demand risk sharing between the shipper and the carrier. The carrier may preposition trucks at a low cost in response to an initial order estimate from the shipper. The shipper finalizes the order and is charged a per-truck penalty for orders above or below a specified percent deviation from the forecast. We characterize the optimal decisions for each party that induce a coordinated channel under a variety of compliance scenarios, and we discuss ways to set the parameters in order to coordinate the channel. In order to aid the implementation of the percent-deviation contract in practice, we develop ways to set the parameters to satisfy both parties' individual-rationality constraints. Since the shipper attains less expected profit than she would under a traditional contract, we consider the discounted wholesale price that the shipper would require to endorse this contract. We also analyze the cost reduction the carrier would need to be willing to offer this discount.

Hui Zou
Department of Statistics
Stanford University

In the practice of statistical modeling, it is often desirable to have an accurate predictive model with a sparse representation. The lasso is a promising model building technique, performing continuous shrinkage and variable selection simultaneously. Although the lasso has shown success in many situations, it may produce unsatisfactory results in some scenarios: (1) the number of predictors (greatly) exceeds the number of observations; (2) the predictors are highly correlated and form

Speaker biograhy:

Xun Yu Zhou (aka XYZ) got his BSc in pure mathematics in 1984 and his PhD in operations research and control theory in 1989, both from Fudan University. He did his postdoctoral researches at Kobe University (Science Faculty) and University of Toronto (Business School) from 1989 to 1993, and joined The Chinese University of Hong Kong (Engineering School) in 1993 where he is now a Professor. His research interests are in stochastic control, financial engineering and discrete-event manufacturing systems. He has published more than 70 journal papers, 1 research monograph, and 2 edited books. He is a Fellow of IEEE.
Other honors include SIAM Outstanding Paper Prize, Croucher Senior Fellowship, and Alexander von Humboldt Research Fellowship. He is or was on the editorial board of Operations Research (1999-), Mathematical Finance (2001-), and IEEE Transactions on Automatic Control (1999-2003).

Analysis of multiple curves was motivated by a genetic microarray
experiment. The data consist of a large number of short expression
profiles over time, which are to be grouped based on their shapes.
This problem is challenging because a large number of profiles are observed over a small number of time points and because the
contrast-to-noise ratio for observations tends to be low. In addition, a fair number of profiles can be constant. To address this and similar problems, a new statistical clustering method based on transformation and smoothing is introduced.

Analysis of multiple peaks was developed in the context of NMR-based protein structure determination. An NMR experiment produces signal data at a large number of time points, which can be investigated in the frequency domain. The primary objective is to identify the frequency peaks, and to estimate their location, width, and height. The challenges in this application are that the signal is inhomogeneous, and some of the peaks are partially or totally overlapping. A new statistical approach based on a finite mixture model of nonlinear regressions is presented.

Drayage orders are generally pickup and delivery requests with time windows. The repositioning of empty equipment may also be required in order to facilitate loaded movements. The drayage orders are satisfied by a heterogeneous fleet of drivers. Driver routes must satisfy various operational constraints.

We present an optimization methodology for finding cost-effective schedules for regional daily drayage operations. The core of the formulation is a set partitioning model whose columns represent routes. Routes are added to the formulation by column generation. We present numerical results for real-world data which demonstrate that our methodology produces low cost solutions in a reasonably short time.

Dr. Smith received his Ph.D. in Engineering Science from the University of California at Berkeley where he held an NSF Fellowship. He holds a bachelors degree in Physics from Harvey Mudd College and an MBA from Berkeley. He is currently the Altarum/ERIM Russell D. O'Neal Professor of Engineering at the University of Michigan, Ann Arbor. He has been a visiting professor at the Statistical Laboratory, University of Cambridge; the Faculty of Systems Theory and Operations Research, Delft University of Technology, The Netherlands; the Department of Industrial Engineering and Operations Research, Columbia University; the Econometric Institute, Erasmus University, Rotterdam, The Netherlands; Laboratoire d'Automatique et d'Analyse du Systems du CNRS, Toulouse, France; the Faculty of Industrial Engineering and Management, Technion, Haifa, Israel; the Mathematical Sciences Research Institute, Berkeley, California; the Operations Research Group, Institute of Pure and Applied Mathematics, Rio de Janeiro, Brazil; the Department of Industrial Engineering and Operations Research, University of California, Berkeley. He is the receipient of the first Altarum/ERIM Russell D. O'Neal Professorship of Engineering at the University of Michigan. He has also been honored with the Distinguished Faculty Achievement Award from the University of Michigan, the College of Engineering Research Excellence Award, an Outstanding Teacher Award from the Michigan Student Assembly, and a National Science Foundation Fellowship. He is a Fellow of the Institute for Operations Research and the Management Sciences. Dr. Smith teaches courses in dynamic programming and stochastic processes. He has supervised the doctoral research of over twenty-one students since 1984. Dr. Smith is Director of the Dynamic Systems Optimization Laboratory at the University of Michigan. The Laboratory research is directed toward the modeling and analysis of dynamical systems over time. He is also the UM Thrust Leader for Manufacturing Systems for the General Motors Collaborative Research Laboratory in Advanced Vehicle Manufacturing at the University of Michigan. Dr. Smith worked earlier at Bell Laboratories in the Network Planning Department where he developed models and algorithms for optimal routing of communications traffic. He is currently project director on three NSF Grants in global and infinite horizon optimization within the Operations Research and Manufacturing Enterprise Systems Programs of the National Science Foundation. Dr. Smith is a Past Associate Editor of Management Science, Chairman of the 2003 Dantzig Dissertation Prize Committee, and is the author of over eighty peer reviewed publications.

To minimize its operating costs, the management company must optimally schedule its available resources to serve all customer demand. In this paper, we analyze several issues important in strategic planning, including the effects of different crew-swapping schedules, demand expansion models, and number of core planes owned by the company. Our analysis requires a solution methodology that takes details in real world situations into consideration. Specifically, our scheduling approach considers: crew transportation and overtime costs, scheduled and unscheduled maintenance effects, crew rules, and the presence of non-crew-compatible fleets.

Computations performed on actual and random data sets show that our algorithm is efficient when compared to other state-of-the-art algorithms.

In the first project "Bidding strategies and their impact on the auctioneer's revenue in combinatorial auctions," focusing on combinatorial auctions, we propose a simple and efficient model for evaluating the value of any bundle given limited information, design bidding strategies that efficiently select desirable bundles, and evaluate the performance of different bundling strategies under various market settings.

In the second project "Retailer shelf-space management in a competitive environment," promotional investment effects are integrated with retail store assortment decisions and shelf space allocation. An optimization model for the category shelf-space allocation incorporating promotion effects is presented. Based on the proposed model, a category shelf space allocation framework with trade allowances is presented where a multi-player Retailer Stackelberg game is introduced to model the interactions between retailer and manufacturers.

In the third project "Supply-chain oriented robust parameter design," we introduce the game theoretical method, commonly used in supply-chain analysis to solve potential conflicts between manufacturers at various stages. These manufacturing chain partners collaboratively decide parameter design settings of the controllable factors to make the product less sensitive to process variations.

Reception at 5:30: Pizza!

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 professionals in the Atlanta community. The meeting will be in the Executive Classroom (Room 228) of the ISyE Main Entrance Bldg. (755 Ferst Drive, formerly the Dupree School of Management Building - same location as our previous recent meetings.) Parking is available next to the building.

http://www.isye.gatech.edu/visitors/maps/

GAMFS, Quasi-Hurst exponents, Mallat Model and Tsallis Maxent
Model are used in the study of assessing the effects of
atmospheric stability on the turbulence measurements in the
inertial subrange. We also formulate the criteria for success in
evaluating how atmospheric stability alters the MFS of a single
flow variable time series as a statistical classification model.
We use the multifractal discriminate model as the solution of this
problem. Also, high frequency pupil-diameter dynamic measurements,
which are well documented as measures of mental workload, are
summarized using both GAMFS and MSM. These summaries are further
used as the feature vector in the Comb-K-NN classifier. The
serious inhomogeneity among subjects in the same user group makes
classification difficult. These difficulties are overcome by using
Comb-K-NN classifier.

In this dissertation, we identified, documented the anatomy of various pull type control policies and compared their performances. Based on the analysis, we designed a new class of control policy called Extended CONWIP Kanban control (ECK). We developed a state space comparison methodology to compare various policies with that of optimum solution. Our analysis reveals that our ECK policy provides the most adjustment flexibility toward the optimum. Our simulation and case study verify the advantages and utilities of the ECK policy. Beside the cost savings, another major advantage of the ECK policy is its robustness. That is, the total cost of the ECK policy is less sensitive to demand or system variations.

The model is a controlled queueing network, where controls have dual
effect. In addition to determining exogenous customer arrival rates,
service rates at the nodes, and (possibly random) routing of customers among the nodes, each control decision produces a certain vector of "commodities." The set of available control choices depends on the underlying random network mode. Network "utility" is a concave
function of the vector of long-term average rates at which commodities are produced. The goal is maximize utility while keeping network queues stable. We introduce a very parsimonious dynamic control policy,
called Greedy Primal-Dual algorithm, and prove its asymptotic optimality. Although the model is formulated in terms
of a queueing network, the algorithm can be viewed as a dynamic mechanism for solving rather general convex optimization problems.

Risk-averse managers value investments that lower portfolio volatility and stabilize returns. This, too, is achieved with assets that provide orthogonal returns: By lowering portfolio volatility and stabilizing returns, these managers improve their portfolios' Sharpe Ratios.

Until recently, both types of managers typically did not think of
commodities as the natural choice to meet both sets of goals vis--vis
risk reduction and orthogonal returns. While this is puzzling given the historical performance of commodities, the industrial practice has been changing.

Finally, we would like to recognize Jiheng Zhang for volunteering to coordinate the weekly tea, and thank the department for supporting our initiative.

If you have any suggestions or comments, please email us at gsac at isye dot gatech dot edu.

Students and faculty are invited to remain after the seminar for a coffee and snacks reception sponsored by the The Logistics Institute.

Regards,
The GSAC (Graduate Student Advisory Committee)

In this thesis, we mainly study the design of the optimal eroding price contract in a two period setting under three different model settings :
(1) Every supplier faces a new cost in each period (NLI model), (2) The supplier who wins the auction in the first period locks-in his cost for the future, and the buyer makes the future payment based on the winning supplier's current bid (LI1), and (3) The supplier who wins the auction in the first period locks-in his cost for the future, and the buyer makes the future payment based on the winning supplier's actual cost (LI2). Under NLI setting, the magnitude of the cost reduction due to learning by doing is common knowledge, while the magnitude is uncertain under LI1 and LI2 settings. We also study the optimal reserve prices in sequential independent auctions under NLI setting.

We go on to compare the performance of the eroding price policy against sequential independent auctions (without or with the optimal reserve
prices) under the above model settings. Via analytical and numerical comparisons, we find that even in the presence of learning by doing/economies of scale in production, a buyer is often better off running sequential auctions with a reserve price, rather than limiting competition and contracting with a single supplier in the hopes of extracting a better future price.

In the second part, optimal designs are identified using Bayesian methods. This work is linked with response surface methodology where the first step is to perform factor screening, followed by response surface exploration using different experiment plans. A Bayesian analysis approach is used that aims to achieve both goals using one experiment design. In addition we use a Bayesian design criterion, based on the priors for the analysis approach. This creates an integrated design and analysis framework. To distinguish between competing models, the HD criterion is used, which is based on the pairwise Hellinger distance between predictive densities.

Mixed-level fractional factorial designs are commonly used in practice but its aliasing relations have not been studied in full rigor. These designs take the form of a product array. Aliasing patterns of mixed level factorial designs are discussed in the third part.

In the fourth part, design of experiment ideas are used to introduce a new global optimization technique called SELC (Sequential Elimination of Level Combinations), which is motivated by genetic algorithms but finds the optimum faster. The two key features of the SELC algorithm, namely, forbidden array and weighted mutation, enhance the performance of the search procedure. Illustration is given with the optimization of three functions, one of which is from Shekel's family. A real example on compound optimization is also given.

Students and faculty are invited to remain after the seminar for a coffee and snacks reception sponsored by the The Logistics Institute.

See File Download for PDF of Abstracts and Bios

---------------------------

:: Data Fusion for Quality Improvements in Complex Solar Cell Manufacturing Processes
First seminar: 9:10 - 9:55 am

Jionghua (Judy) Jin

Assistant Professor

Department of Systems and Industrial Engineering

University of Arizona

---------------------------

:: Analysis and Design Methods of Distributed Sensor Systems for Manufacturing Quality Improvement

Second Seminar: 10:05 - 10:50 am

Yu Ding

Assistant Professor

Dept. of Industrial Engineering

Texas A&M University

---------------------------

:: Recent Advances in Statistical Quality Control

Third Seminar: 10:55 - 11:40 am

Fugee Tsung

Professor

Dept. of Industrial Eng & Eng Mgt

HK Univ. of Science & Technology

---------------------------

Note: If you would like to meet some of these seminar speakers in the afternoon (2-4 pm) of May 17th, please let me know before Tuesday noon, May 10th.

This talk is based on a joint work with Inchi Hu.

Part II: Comparison of due-date oriented dispatching rules.
In the second part of the talk we compare the cycle-time and on-time-delivery performance of the Critical Ratio (CR), the Earliest Due Date (EDD), and the Operation Due Date (ODD) dispatching rules in semiconductor wafer manufacturing. In addition, we comment on the robustness of the rules to low target flow-factor values. It turns out that EDD has almost no positive effect on on-time delivery compared to First-In-First-Out (FIFO) dispatching. CR and ODD, however, reduce the tardiness of lots considerably. ODD dispatching is more robust than CR if the target cycle times for the factory are chosen smaller than the corresponding average cycle times of a factory under FIFO dispatching regime.

Part III: Accelerating products for due-date oriented dispatching. In semiconductor manufacturing facilities, there is often the need to speed up certain product types. This is usually done by either assigning higher priorities or by reducing due dates. Here, we study the effects of accelerating one product type by a tighter due date on the on-time delivery performance of the other products manufactured. It turns out that the results depend on the considered factory, its load, and the accelerated product. As a consequence, it will be hard for production planners to find simple rules of thumb for the effects of accelerating products. In general, detailed simulation experiments will be required.

Biographical sketch:

OLIVER ROSE is Professor for Modeling and Simulation in the Department of Computer Science at the Dresden University of Technology, Germany. He received an M.S. degree in applied mathematics (1992) and a Ph.D. degree in computer science (1997) from the University of W

truss topolology design problems in mechanical/civil engineering

signal processing problems in electronic engineering.

In each case we reveal some "hidden convexity," enabeling us to convert the original problems into tractable convex programming problems.

Students and faculty are invited to remain after the seminar for a coffee and snacks reception sponsored by the The Logistics Institute.

Motivated by the possibility of port of entry closures in the event of a security incident, we specialize the previous model to a two-stage international supply chain. A domestic manufacturer orders a single product from a foreign supplier and the orders must cross an international border that is subject to closure. We first assume that border congestion is negligible. The manufacturer's optimal inventory policy and long-run average cost are analyzed. We present structural policy results and the results of a comprehensive numerical study that have important implications for business and for the cooperation between business and government in disruption management and contingency planning.

Finally we extend the border closure model to include both border closures and the resulting congestion. We model the border processing system as a discrete-time, single-server queue with constant arrival rate and Markov-modulated service rate. A key task is the development of the leadtime distribution, which is more complex than in the previous model. We present the results of a comprehensive numerical study and provide managerial insights.

Students and faculty are invited to remain after the seminar for a coffee and snacks reception sponsored by The Logistics Institute.

Although SDPs are polynomially solvable, it is still very challenging to solve large-scale SDPs efficiently. The second topic of this thesis is to provide an approach for solving large-scale well-structured sparse SDPs via a saddle point mirror-prox algorithm with ${cal O}(epsilon^{-1})$ efficiency by exploiting sparsity structure and reformulating SDPs into smooth convex-concave saddle point problems.

The third topic of this thesis is to develop a long-step primal-dual infeasible path-following algorithm for convex quadratic programming
(CQP) whose search directions are computed by means of a preconditioned iterative linear solver. A uniform bound, depending only on the CQP data, on the number of iterations needed by a preconditioned iterative linear solver is established. A polynomial bound on the number of iterations of this method is also obtained.

The last topic of this thesis is to develop an efficient ``nearly exact''
type of method for solving large-scale ``low-rank'' trust region subproblems by completely avoiding the computations of Cholesky or partial Cholesky factorizations. We also provide a computational study on this method by applying it to solve some large-scale nonlinear programming problems.

The models we develop are space-time measure-valued Markov processes. Specifically, particles enter a space according to a space-time Poisson process and are assigned independent and identically distributed attributes. The attributes may determine their movement in the space, and whenever a new particle arrives, it randomly deletes particles from the system according to their attributes.

Our main result establishes that spatial Poisson processes are natural temporal limits for a large class of particle systems.
Other results include the probability distributions of the sojourn times of particles in the systems, and probabilities of numbers of customers in spatial polling systems without Poisson limits.

In the first part of this dissertation, we investigate the problem of the detectability of an inhomogeneous convex region in a Gaussian random field. The first proposed detection method relies on checking a constructed statistic on each convex set within an n

Students and faculty are invited to remain after the seminar for a coffee and snacks reception sponsored by The Logistics Institute.

The empirical work studies the cause of low aggregation of timber
supply in reverse auctions of an online timber exchange. Unlike
previous research results regarding timber auctions, which focus on
offline public auctions held by the U.S. Forest Service, we study
online private auctions between logging companies and mills. A
limited survey of the online auction data revealed that the auctions
were successful less than $50%$ of the time. Regression analysis is
used to determine which internal and external factors to the auction
affect the aggregation of timber in an effort to determine the
reason that so few auctions succeeded. The analysis revealed that
the number of bidders, the description of the good, and the volume
demanded had a significant influence on the amount of timber
supplied through the online auction exchange. A plausible
explanation for the low aggregation is that the exchange was better
suited to check the availability for custom cuts of timber and to
transact standard timber.

By specializing to systems such as the M/M/1 queue with time varying rates, we show
that the natural scale invariance property for the processor sharing service discipline corresponds
directly to the uniform acceleration scaling limits necessary for its queueing fluid and diffusion limits..

We use these asymptotic results to obtain strong law of large numbers and central limit theorems
for the virtual sojourn time of any given job size for the M/M/1 queue with time varying rates. This
gives us simple fluid and diffusion approximations for the sojourn times of virtual customers.

This is joint research with Mor Harchol-Balter of Carnegie Mellon University and Robert C.Hampshire
of Princeton University.

The model is shown to be a consistent method for estimating the parameters
of the optimal solution to the problem of maximizing the probability of
correct classification subject to limits on inter-group misclassification
probabilities. A polynomial time algorithm is described for two-group
instances. The method is NP-complete for a general number of
groups, and an approximation is formulated as a mixed-integer program
(MIP). The MIP is difficult to solve due to the formulation of
constraints wherein certain variables are equal to the maximum of a set of
linear functions. These constraints are conducive to an ill-conditioned
coefficient matrix. Methods for generating edges of the conflict graph
and conflict hypergraphs are discussed. The conflict graph is employed
for finding cuts in a branch-and-bound framework. This technique and
others lead to improvement in solution time over industry-standard
software on instances generated by real-world data. The classification
accuracy of the model in relation to standard classification methods on
real-world and simulated data is also noted.

This dissertation includes three major parts. The first part develops a prototype model for a decentralized reverse production system with two tiers, collectors and processors, focusing on the coordination of transactions of recycled items between these two tiers. The collectors determine the individual material flow allocation mechanisms that relate the flow amount to the overall vector of acquisition prices that are determined by the processors. The processors compete for the flow from the collectors and reach an equilibrium state where no entity is willing to change its decisions.

In the second part, we extend the prototype model for a general multitiered recycling network comprised of the upstream boundary tier, several intermediate tiers, and the downstream boundary tier where each of tiers is with multiple independent entities. Recycled items flow from the top tier to the downstream tier but acquisition prices are set from the downstream tier back to the upstream tier.

Finally the third part provides a comparison of centralized and decentralized models for reverse production systems and addresses several numerical insights of different system subsidy schemes.

We first consider the situation where the forward area includes all SKUs so that order picking is performed only in the forward area. In this case, the problem can be decomposed. The resulting sub-problem is convex and can be solved very efficiently based on the Karush-Kuhn-Tucker (KKT) conditions. This property enables the use of a Generalized Benders Decomposition (GBD) method to solve the sizing and dimensioning problem exactly.

We then extend the problem to more general situations where the forward area contains a subset of SKUs. This requires integrating the sizing and dimensioning decisions with the decision to assign SKUs to the forward area based on their flow characteristics (i.e., the forward reserve allocation). A similar decomposition strategy can be employed, but the sub-problem (incorporating the forward reserve allocation) is no longer convex. A bi-level hierarchical heuristic approach is proposed that integrates a pattern search method for the master problem and optimal and heuristic algorithms for the sub-problems.

Numerical results demonstrate that the proposed solution methods can efficiently find optimal or near optimal solutions for the sizing and dimensioning problem, and the resulting solutions are robust with regards to possible forecasting errors in design parameters.

An important decision for a carrier is to determine the prices specified in those contracts. An ocean carrier would like to structure the prices to maximize its total revenue while obtaining best match between supply and demand. In this study, we propose optimization models, which can be used as decision tools by ocean carriers before or during the negotiation seasons. Two deterministic models are formulated and capture the operational details at different levels. To handle uncertainties, a stochastic model is developed. The three-stage problem is solved by bundle method and trust region method. Output of the models includes optimal prices, anticipated cargo flows and equipment flows. Such output of the models can help ocean carriers to develop their negotiation strategies and cargo plans for the coming season.

Students and faculty are invited to remain after the seminar for a coffee and snacks reception until 5:30 pm.

Prior to founding Blackstone, Peterson was chair and CEO of Lehman Brothers (1973-77), and later held that same position with Lehman Brothers, Kuhn, Loeb, Inc. (1977-84). He was chair and CEO of Bell and Howell Corporation from 1963-1971. In 1971, President Richard Nixon named Peterson assistant to the president for International Economic Affairs. In 1972, he was similarly appointed U.S. Secretary of Commerce and assumed the chairmanship of Nixon

Investment by individual supply chain members would help ensure security, although contribution to this public good is not always in their best interest. We investigate the security of a general supply chain and study the interactions of the members of the chain and how these interactions affect investment decisions in security. We design incentive mechanisms such as the sharing of destruction costs, price differentiation, and the allocation of public catastrophe costs to encourage supply chain members to invest in security. We also demonstrate that competition between suppliers can enhance public security, whereas competition between retailers can harm it.

Students and faculty are invited to remain after the seminar for a coffee and snacks reception sponsored by the The Logistics Institute.

Modeling electricity spot prices is essential for asset and project valuation as well as risk management. I introduce the mean-reversion feature into a classical variance gamma model to model the electricity price dynamics. The generalized method of moments and the Markov Chain Monte Carlo method are proposed to estimate parameters. Derivatives pricing formulae are derived through transform analysis.

Under a realistic electricity price model with mean-reversion and jumps, pricing and hedging customized power contracts is very challenging. A tolling agreement (or tolling contact) is one example of customized power contract in which a contract buyer reserves the right to take the output of an underlying electricity generation asset by paying a predetermined premium to the asset owner. I propose a real options approach to value a tolling contract incorporating operational characteristics of the generation asset and contractual constraints. Two simulation-based methods are proposed to solve the valuation problem. The effects of different electricity price assumptions on the valuation of tolling contracts are examined. Based on the valuation model, I also propose a heuristic scheme for hedging tolling contracts and demonstrate the validity of the hedging scheme through numerical examples.

An estimator of conditional value-at-risk (VaR) based on a GARCH(1, 1) model with heavy tailed innovations is proposed. The normal approximation method and a data tilting method are proposed to constructe confidence intervals for the conditional VaR estimator and assess their accuracies by simulation studies.

We present some results of ongoing work that is motivated by the question whether or not such bounds also can be obtained by applying the more elegant theory of self-concordant functions to barrier functions based on kernel functions. A major difficulty that arises when dealing with this question is that in general these barrier functions are not self-concordant. As we will show, however, on the central path and in its neighborhood they behave as being self-concordant; we call them locally self-concordant. As a consequence we expect it to be possible to answer the above question positively. In this talk we restrict ourselves to a special case of small-update methods, namely to full-Newton step methods.

For these schemes we prove general convergence results. We establish also the global and local worst-case complexity bounds. It is shown that the corresponding search directions can be computed by a standard linear algebra technique. We discuss also a possibility to accelerate the scheme on the convex problems.

---------
Abstract
---------
In this talk we explore the problem of determining the optimal inspection/disposition policy for a finite batch of items produced by a machine which is subject to random breakdowns. This talk will explore two separate, but related works.

In the first work we apply information theory to the problem of finding the first nonconforming unit in the batch at minimum cost. Two distinct but related aspects of this problem are treated: determining which units should be inspected and determining how many units should be sent for inspection at the same time. The solution is based on the principles of inspecting the units that maximize the reduction in the uncertainty regarding the location of the first nonconforming unit and of minimizing the cost per unit of uncertainty reduced. These principles are formalized by means of a series of theorems leading to an easy-to-implement algorithm for managing parallel inspection. This approach is successfully compared to the optimal solution obtained with dynamic programming and to other heuristics.

In the second work, we investigate situations in which the disposition of some of the units can be made without definitely identifying their status. The model considers three costs: the cost of inspection, the cost of rejecting a conforming unit, and the cost of accepting a nonconforming unit. We identify which units should be inspected and in which order so as to minimize costs. We place special emphasis on three different policies: the cost minimizing policy; the policy of perfect information, i.e. we insist on determining the quality of each unit; and the policy of zero-defects, i.e. we insist that all accepted units are known to conform to specifications, allowing the rejection of units of unknown quality.

We hope to see you there!

Your Graduate Student Advisory Committee
Clarence, Dylan, Lori, & Zhiguang

Students and faculty are invited to remain after the seminar for a coffee and snacks reception sponsored by The Logistics Institute.

Wavelet-based methods are important in statistics in areas such as regression, density and function estimation, factor analysis, modeling and forecasting in time series analysis, assessing self-similarity and fractality in data, and spatial statistics. In this thesis we show applicability of the wavelets by considering three problems:

First, we consider a binary wavelet-based linear classifier. Both consistency results and implemental issues are addressed. We show that under mild assumptions wavelet-based classification rule is both weakly and strongly universally consistent. The proposed method is illustrated on synthetic data sets in which the "truth" is known and on applied classification problems from the industrial and bioengineering fields.

Second, we develop wavelet shrinkage methodology based on testing multiple hypotheses in the wavelet domain. The shrinkage/thresholding approach by implicit or explicit simultaneous testing of many hypotheses had been considered by many researchers and goes back to the early 1990's. We propose two new approaches to wavelet shrinkage/thresholding based on local False Discovery Rate (FDR), Bayes factors and ordering of posterior probabilities.

Finally, we propose a novel method for the analysis of straight-line alignment of features in the images based on Hough and Wavelet transforms. The new method is designed to work specifically with Transmission Electron Microscope (TEM) images taken at nanoscale to detect linear structure formed by the atomic lattice.

We analytically characterize the effects of the project's characteristics---its systematic and technical risk, and the degree of asymmetry in beliefs about its quality---on the path of staged investments by the VC and the structure of the long-term contract between the VC and the EN. Consistent with empirical evidence, we predict that varying project characteristics lead to significant heterogeneity in contractual structures and investment schedules. The systematic and technical risks of projects have opposite effects on the durations and economic values of VC relationships. The duration, firm value, and the expected payoff to the VC decrease with the project's systematic risk but increase with its technical risk, which leads to the striking implication that the value of the firm and the expected payoff to the VC are actually enhanced when there is greater noise in the perception of project quality.

Broadly, our study not only demonstrates that the interactions among agency conflicts, imperfect information, and asymmetric beliefs have a major impact on the VC-EN relationship, but also precisely describes the manner in which they affect this relationship.

Partly as a result of this, a number of different models have been considered for analyzing data from a crossover design. In this talk we will focus on the optimal design question. Thus, given that we want to use a certain model, what is a good design to use? We will restrict our attention to the situation that only two treatments are to be compared in the experiment.

Students and faculty are invited to remain after the seminar for a coffee and snacks reception sponsored by The Logistics Institute.

Colonel Visker, a native of Albany, NY, was commissioned a Second Lieutenant in the Quartermaster Corps upon graduation from Siena College, where he was a distinguished military graduate in 1979. At the same time he was awarded a bachelors degree in Banking and Finance. He also holds a Masters of Science in Systems Management (Logistics) from the University of Southern California and a Masters of Science in Strategy from the US Army War College.

Prior to his arrival at the CENTCOM DDOC and the San Joaquin Distribution Center, Colonel Visker commanded the 55th Theater Material Management Center at Camp Henry, South Korea. In other previous assignments he commanded Defense Distribution Depot Anniston, AL; held joint duty assignments with Special Operations Command Central (Airborne), MacDill AFB, FL where he was the Headquarters Commandant and Director of Logistics, J4. At Fort Bragg, NC; he was the Chief of Logistics Plans with the 1st Corps Support Command (Airborne) and Executive Officer of the 264th Corps Support Battalion. Colonel Visker also served as the Supervising Contracting Officer for the Defense Logistics Agency