 |
|
||||||||
Plenary Lectures
| Dimitris J. Bertsimas | A computationally tractable theory of performance analysis in stochastic systems | |
|
Co-director, Operations Research Center, MIT, Cambridge  |
Modern probability theory, whose foundation is based on the axioms set forth by Kolmogorov, is currently the major tool for performance analysis in stochastic systems. While it offers insights in understanding such systems, probability theory is really not a computationally tractable theory. Correspondingly, some of its major areas of application remain unsolved when the underlying systems become multidimensional: Queueing networks, network information theory, pricing multi-dimensional financial contracts, auction design in multi-item, multi-bidder auctions among others. We propose a new approach to analyze stochastic systems based on robust optimization. The key idea is to replace the Kolmogorov axioms as primitives of probability theory, with some of the asymptotic implications of probability theory: the central limit theorem and law of large numbers and to define appropriate robust optimization problems to perform performance analysis. In this way, the performance analysis questions become highly structured optimization problems (linear, conic, mixed integer) for which there exist efficient, practical algorithms that are capable of solving truly large scale systems. We demonstrate that the proposed approach achieves a computationally tractable methods for (a) analyzing multiclass queueing networks, (b) characterizing the capacity region of network information theory and associated coding and decoding methods generalizing the work of Shannon, (c) pricing multi-dimensional financial contracts generalizing the work of Black, Scholes and Merton, (d) designing multi-item, multi-bidder auctions generalizing the work of Myerson. This is joint work with my doctoral student at MIT Chaitanya Bandi. |
|
| Kenneth L. Judd | Numerically Efficient and Stable Algorithms for Solving Large Dynamic Programming Problems in Economics, Finance, and Climate Change Models | |
Hoover Institution, Stanford |
Life is a very large dynamic programming problem. The standard approach to analyzing real problems is to formulate highly stylized and unreliable simplifications, and use analytical tools or primitive numerical methods. This is no longer necessary with the computational tools that are now available. We are developing DPSOL, a flexible and robust set of computational tools for discrete-time dynamic programming based on modern computational methods. We use multivariate approximation methods to represent value functions. Standard approximation methods, such as least squares curve fitting, often lead to numerically unstable iterations. We stabilize value function iteration by using shape-preserving approximation methods. We also use Hermite interpolation to improve the quality of the approximation with only minor extra cost. We use efficient multivariate quadrature methods to compute expectations, and modern optimization methods to solve the Bellman equations. We combine these ideas to develop a nonlinear programming approach to dynamic programming. We have developed parallel versions to exploit distributed computing systems. This is illustrated with applications such as (a) portfolio choice with six stocks, one bond, and transaction costs, (b) life-cycle decisions over consumption, investment, labor supply, and education choices, (c) a dynamic stochastic generalization of DICE, a well-known model used in climate change studies. More generally, we lay out a conceptual framework for thinking about numerical solutions of dynamic programming problems and incorporating past and future advances in algorithms and hardware. This is joint work with Yongyang Cai, a postdoctoral student at the Hoover Institution. |
|
| William R. Pulleyblank | Optimizing Twenty-First Century Decision Making | |
|
Professor, Operations Research, Mathematical Sciences Department, United States Military Academy, West Point, NY  |
For the last decade, there has been increasing recognition of the importance of advanced analytics and optimization methods to public and private organizations. This is being driven by increased global competition as well as an accelerated pace of business. It coincides with the emergence of new hardware and software paradigms as well as the capability to integrate advanced analytic methods in our planning and operational activities. Here are some of the main issues: 1. How do we deal with unprecedented amounts of data in our decision making? 2. How can we effectively incorporate risk measurement and management in our decision making? 3. How do we enable the transition from strategic planning systems to real-time operational systems? 4. What optimization and analytics opportunities are being created by the rapid growth of social networks? From 2005 to 2009, as part of IBM’s Global Business Services, I led an organization with the mission of developing and delivering these capabilities to clients in all sectors. This has subsequently become a major part of IBM’s consulting business. Since 2010 I have been a member of the faculty at The United States Military Academy, West Point and have discovered that top-of-mind issues in the military are very similar. I will discuss challenges and issues encountered, as well as examples of successful projects dealing with these topics. These include steel mill scheduling, optimizing call center profitability and dealing with uncertainty related to government policy decisions. I will also discuss obstacles that remain to be overcome. |
|
Semi-plenary Lectures
|
David Alderson |
Robustness, Design, and Complexity and Their Implications For Network-Centric Infrastructures |
|
|
Naval Postgraduate School, Monterey
Topic: Network Industries and Regulation  |
Many cyber-technical visions convincingly suggest that network-centric technology will provide unprecedented levels of capability and efficiency to support the operation and management of modern society’s most vital functions—ranging from delivery of economic goods and services, business processes, global financial markets, education, health care, military operations and national defense, and other government services. A fundamental challenge is to understand and manage the growing complexity of these systems. In this talk, I will revisit the notion of “organized complexity” and suggest that it is fundamental to what we define as essential in these network systems. I argue that complexity arises in highly evolved technological (and biological) systems primarily to provide mechanisms to create robustness. This view of complexity in highly organized systems is fundamentally different from the dominant perspective in the field of network science, which downplays function, constraints, and tradeoffs, and tends to minimize the role of design. |
|
|
Sally Brailsford |
OVERCOMING THE BARRIERS: GETTING SIMULATION USED IN HEALTHCARE |
|
|
University of Southampton Topic: Health, life sciences, bioinformatics  |
This talk addresses a key, controversial issue in the health OR literature, namely the apparent failure of OR modelling in general and simulation in particular to become embedded and widely implemented as practical management tools within healthcare organizations. There is a massive academic literature in this field, but the vast majority of published papers are either purely theoretical or report individual one-off success stories. The evidence suggests that simulation has failed to become part of the regular “management toolkit” in the healthcare sector, in contrast with its success in manufacturing and service industries, and the military and defence sectors. The reasons for this remain unclear. The research presented here is a case study to evaluate the adoption (or otherwise) of one particular simulation modelling tool, Scenario Generator, which was developed by the SIMUL8 Corporation in a collaborative partnership with the UK’s National Health Service Institute for Innovation and Improvement. The software was offered free of charge for one year to any UK healthcare organisation that wished to try it. The deal also included two full days of face-to-face hands-on training plus unlimited telephone support. Our study involved semi-structured interviews with employees of 28 organisations who had all been engaged in some way with the initiative. In addition to interviewing active users, we also talked to people who had tried the software but had given up, and people who had decided not to use it. The latter two groups provided some particularly useful insights. In this talk we present a brief summary of barriers and facilitators to the successful use of the Scenario Generator software itself, but the main aim is to focus more broadly on factors influencing the successful adoption of simulation tools in general within healthcare organisations. The insights gained in this case study are relevant to improving the uptake of OR modelling in general within healthcare organisations anywhere in the world. |
|
|
Peter Bühlmann |
Statistics and Optimization for Causal Inference in Large-Scale Biological Systems |
|
|
ETH Zurich
Topic: Statistics, computational biology  |
Understanding cause-effect relationships between variables is of great interest in many fields of science. An ambitious but highly desirable goal is to infer causal effects from observational data obtained by observing a system of interest without subjecting it to interventions, thereby circumventing severe experimental constraints or exhibiting much lower costs. This is particularly relevant for many important questions in molecular biology. We discuss recent progress that has been achieved over the last few years in statistical graphical modeling and optimization for causal inference, particularly in high-dimensional, sparse settings with thousands of variables but based on only a few dozens of observations. We highlight exciting possibilities, fundamental limitations of any modeling approach, and we discuss two successful experimental validations in the context of molecular biology for yeast (Saccharomyces cerevisiae) and the model plant Arabidopsis thaliana. |
|
|
Daniel Costa |
Decision Support Models in Supply Chain at Nestlé |
|
|
Nestlé Suisse S.A.
Topic: OR in industry and 50 years of SVOR  |
This lecture will start with an overview of the supply chain in Nestlé, from the purchasing of raw and packing materials to the final delivery point at our customers. Then we will focus on some applications where simple decision support and optimization tools have proven very helpful over the years. Such applications include : building of an efficient pallet or container load, simulation of picking activities in a warehouse, identification of optimal warehouse locations, calculation of safety stocks, search for an optimum product portfolio. |
|
|
Paul Embrechts |
The modelling of rare events: from methodology to practice and back |
|
|
ETH Zurich
Topic: Financial modeling, risk management, banking  |
The current discussion around the 2011 Earthquake in Japan, and the 2007-09 Financial Crisis have once more brought rare, but extreme events to the forefront of the public debate. In this talk I will review some of the methodology underlying modern Extreme Value Theory (EVT), show where EVT has practical value, warn for areas where its applicability is dubious and discuss where more research is needed.
The talk is very much based on scientific papers to be found on my website www.math.ethz.ch/~embrechts, and in particular in the textbooks: |
|
|
Karl Isler |
Developments in Booking Control Optimization by Airline Revenue Management |
|
|
Swiss International Air Lines
Topic: Revenue management and dynamic pricing  |
Price differentiation in the airline industry leads to significant revenue opportunities realized by efficiently controlling bookings for the various price products. We give an overview of common practices in revenue management from single flight control to network optimization and dynamic pricing. We discuss the critical role of distribution technology advancements from the traditional travel agent to Internet distribution. |
|
|
Sven Leyffer |
MINOTAUR: Solving Mixed-Integer and Nonlinear Optimization Problems |
|
|
Argonne National Laboratory
Topic: Continuous optimization and control  |
Scientists and engineers are increasingly turning from the simulation of complex processes to the optimization and design of complex systems. Many important design problems involve not only continuous variables with nonlinear relationships but also discrete decisions, giving rise
to mixed-integer nonlinear programming problems (MINLPs). MINLPs combine the combinatorial complexity of the discrete decisions with the numerical challenges of the nonlinear functions.
We present a new package for solving mixed-integer nonlinear optimization problems, called MINOTAUR. The MINOTAUR toolkit is designed to provide a flexible and efficient framework for solving MINLPs. The code s developed in a modular way to enable developers and users to efficiently combine the knowledge of problem structure with algorithmic insights. We will survey recent developments in MINLP and present the underlying algorithmic ideas of MINOTAUR. Our talk will focus on the integration of nonlinear solvers into the MINOTAUR's branch-and-cut framework, and highlight challenges and opportunities for nonlinear optimization. |
|
|
Todd Munson |
CIM-EARTH: Overview and Case Study Co-Authors: Joshua Elliott, Ian Foster, Margaret Loudermilk |
|
|
Argonne National Laboratory
Topic: Energy, environment and climate  |
CIM-EARTH is a framework for specifying and solving computable general equilibrium models. In this talk, I will give an overview of the state of the framework and then discuss a case study using this framework on the international trade implications of carbon policies and border tax adjustments. Carbon contents for the border tax adjustments are computed endogenous to the model by applying a carbon conservation principle. Results are presented using matrices of bilateral carbon flows to analyze the amount of leakage resulting from the policies studied. I will also provide some insights into the sensitivities of the outcomes to uncertain parameters based on the results from a large-scale computational study. |
|
|
Yurii Nesterov |
Random methods in Convex Minimization |
|
|
Université catholique de Louvain
Topic: Convex optimization  |
In this talk, we present the recent results on the complexity bounds for methods of Convex Optimization based only on computation of the function values or directional derivatives. First, we analyze the behavior of the random coordinate descent method and show that for some problem classes it can be more efficient than the usual gradient scheme. In the second part of the talk we discuss pure random search strategies based on normally distributed random Gaussian vectors. Such methods usually need at most $n$ times more iterations than the standard gradient methods, where $n$ is the dimension of the space of variables. This conclusion is true both for nonsmooth and smooth problems. For the later class, we develop also an accelerated scheme with the expected rate of convergence $O({n^2 \over k^2})$, where $k$ is the iteration counter. For Stochastic Optimization, we propose a zero-order scheme and justify its expected rate of convergence $O({n \over k^{1/2}})$. We give also some bounds for the rate of convergence of the random gradient-free methods to stationary points of nonconvex functions, both for smooth and nonsmooth cases. Our theoretical results are supported by preliminary computational experiments. |
|
|
Gernot Tragler |
Operations Research, Drugs, and Optimal Control |
|
|
TU Vienna
Topic: Drugs: Application of optimal control  |
For several decades, Pontryagin’s maximum principle has been applied to solve optimal control problems in engineering, economics, or management. Early Operations Research applications of optimal control include problems such as production planning, inventory control, maintenance, marketing, or pollution control. Since the mid-nineties, optimal control models of illicit drug consumption have contributed successfully to a better understanding of drug epidemics and their control via an optimal mix of instruments such as prevention, treatment, or law enforcement. This talk explains why and how tools of dynamic optimization are used to address pressing questions arising in drug policy. Moreover, methodological advances in optimal control theory that have been triggered by solving these problems will be highlighted (e.g., multiple equilibria & thresholds, age-structured control systems, numerical analysis). |
|
|
Rudolf Vetschera |
Analyzing e-negotiations from a quantitative and qualitative perspective |
|
|
University of Vienna
Topic: Decision analysis and negotiations  |
We give a broad overview of a set of studies on electronic negotiation processes. Our approach combines an analysis of the substantive issues of negotiations, in particular offers, and the communication and relationship aspects. At both levels, we focus on the negotiation process, its influence factors, and the resulting outcomes of negotiations. we show how context factors, like negotiator's preferences, support systems used or characteristics of the negotiation problem influence both quantitative and qualitative aspects of the negotiation process, like concession levels or the structure of communication acts. These process characteristics in turn can be demonstrated to impact outcomes. By various examples of such quantitative and qualitative studies, we show how an integration of different aspects can lead to a comprehensive understanding of negotiation processes. |
|
|
Stefan Voss |
MATHEURISTICS: Hybridizing Metaheuristics and Mathematical Programming |
|
|
University of Hamburg
Topic: Metaheuristics and biologically inspired approaches  |
The field of metaheuristics has traditionally been very receptive to proposals about how to structure algorithms in order to effectively solve optimization problems. Innovation of solution approaches has always been one of the traits of the field, and design paradigms have succeeded as inspiration for algorithm designers: inspiration from nature, improvements of local search, logics and probability, etc. In this presentation we aim to show how both metaheuristics and mathematical programming (MP) can leverage on one another. This especially relates to the term Matheuristics, which describes works that are along these lines, e.g., exploiting MP techniques in (meta)heuristic frameworks or on granting to MP approaches the cross-problem robustness and constrained CPU-time effectiveness which may characterize metaheuristics. This follows a trend in hybridization, which appeared in several forms in the last years: metaheuristics are being hybridized with artificial intelligence, with constraint programming, with statistics, not to mention among themselves. However, the combination of metaheuristics and MP has a set-apart condition. Including MP techniques for a metaheuristic designer does not mean looking for contributions, which could possibly derive from another research area, it means looking inside one’s own cultural baggage, it means using in a different way something one has already had experience with. Examples and applications of matheuristics in various problem domains are surveyed, including logistics as well as telecommunications. |
|
|
Robert Weismantel |
A Cutting Plane Theory for Mixed Integer Optimization |
|
|
ETH Zurich
Topic: Discrete optimization, graphs & networks  |
From a practical perspective, mixed integer optimization represents a very powerful modeling paradigm. Its modeling power, however, comes with a price. The presence of both integer and continuous variables results in an increase in complexity over the pure integer case with respect to geometric, algebraic, combinatorial and algorithmic properties. Specifically, the theory of cutting planes for mixed integer linear optimization has not yet been at a similar level of development as in the pure integer case. The goal of this talk is to discuss four research directions that are expected to contribute to the development of this field of optimization. In particular, we examine a new geometric approach based on lattice point free polyhedra and use it for developing a cutting plane theory for mixed integer sets. We expect that these novel developments will shed some light on the additional complexity that goes along with mixing discrete and continuous variables. |
|
Wichtiger Hinweis:
Diese Website wird in älteren Versionen von Netscape ohne
graphische Elemente dargestellt. Die Funktionalität der
Website ist aber trotzdem gewährleistet. Wenn Sie diese
Website regelmässig benutzen, empfehlen wir Ihnen, auf
Ihrem Computer einen aktuellen Browser zu installieren. Weitere
Informationen finden Sie auf
folgender
Seite.
Important Note:
The content in this site is accessible to any browser or
Internet device, however, some graphics will display correctly
only in the newer versions of Netscape. To get the most out of
our site we suggest you upgrade to a newer browser.
More
information
