wiki:JJHFCOIN10

INFORMS 2010 - a COIN-OR double celebration!

This year marks the 10th anniversary of COIN-OR, and coincidentally, the "retirement" of John Forrest - the founding COIN-OR member, and the one who holds the record of the most bug contributions!

The celebration will be held at the Austin 2010 INFORMS Annual Meeting.

Reception Sunday Night

There will be a thoroughly optimized reception Sunday night, sponsored by IBM and others, at the Iron Cactus from 8-10pm.

Special Invited Cluster

Robin Lougee and Alan King are organizing a special invited cluster, co-sponsored by the INFORMS Computing Society. Don't miss the first invited cluster in INFORMS history that uses the "bitwise inclusive or" operator in its title!! Contact Robin rlougee at us ibm com or Alan kingaj at us ibm com.

  • Confirmed Speakers (in program order)
    • Robin Lougee, Program Manager, COIN-OR, IBM Research
    • Alan King, IBM Research, Joao Goncalves,IBM Research
    • Kipp Martin, University of Chicago
    • Imre Polik, Lehigh University
    • Stuart Mitchell, Research Fellow, University of Auckland
    • Mirko Maischberger, Università degli Studi di Firenze
    • Leo Lopes, Monash University Maths, Australia
    • Kate Smith-Miles, Professor, School of Mathematical Sciences, Monash University
    • Andrew Mason, Dept of Engineering Science, University of Auckland
    • Geoff Leyland, Director, Incremental Limited
    • Michelle Liu, Assistant Professor of IT, Marymount University,
    • John Tomlin, Principal Research Scientist, Yahoo! Research
    • John Forrest, IBM Research (Retd.)
    • David Jensen, IBM Research
    • J.P. Fasano, IBM Research Watson
    • Katya Scheinberg, Research Sceintist, Columbia University,
    • Donald Goldfarb, Professor, Columbia University
    • Zaiwen Wen, NSF Math Institutes' postdoc
    • Zhiwei (Tony) Qin, Columbia University
    • Donald Goldfarb, Professor, Columbia University
    • Pietro Belotti, Clemson University,
    • Giacomo Nannicini, Carnegie Mellon University
    • Jim Ostrowski, jostrows@…
    • Leo Liberti, Associate Professor, CNRS LIX, Ecole Polytechnique,
    • Andrea Qualizza, Carnegie Mellon University
    • Francois Margot, Professor, Carnegie Mellon University
    • William Hart, Sandia National Laboratories
    • Jean-Paul Watson, Sandia National Laboratories
    • David Woodruff, Professor, University of California Davis
    • Daniel Word, Texas A&M University
    • Carl Laird, Texas A&M University
    • John Siirola,S andia National Laboratories
    • Gabe Hackebeil, Texas A&M University
    • Laszlo Ladanyi, CPLEX R&D
    • Philip Walton, Engineering Manager, JBHunt Transport, Inc.
    • Anju Kambadur, IBM T J Watson Research Center
    • Matthew Galati, Optimization Interface Lead, SAS Institute
    • Ted Ralphs, Associate Professor, Lehigh University
    • Ellis Johnson, Professor, Georgia Tech
    • Feng Qiu, Student, Georgia Tech
    • Shabbir Ahmed, Associate Professor, Georgia Institute of Tech
    • Santanu Dey, Assistant Professor, Georgia Institute of Technology
    • Yu-Heng Chang, Georgia Institute of Technology
    • John-Paul Clarke, Georgia Institute of Technology
    • Senay Solak, University of Massachusetts Amherst
    • Earl Barnes, Georgia Tech, Atlanta GA
    • John-Paul Clarke, Georgia Institute of Technology
    • Gerd Infanger, Consulting Professor, Stanford University
    • John Birge, University of Chicago
    • Dharmashankar Subramanian,R esearch Staff Member, IBM Research
    • Pu Huang, Research Staff Member, IBM Research
    • Gyana Parija, IBM India Research Lab
    • Matthew Saltzman, Clemson University
    • Bjarni Kristjansson, President, Maximal Software, Inc
    • Steven Dirkse, Director of Optimization, GAMS Development Corp.
    • Michael Bussieck, GAMS Development Corp.
    • Jan Jagla, GAMS Development Corp.
    • Alex Meeraus, President, GAMS Development Corp.
    • Paul Vandereijk, GAMS Development Corp.
    • Lutz Westermann, GAMS Development Corp
    • David M. Gay, AMPL Optimization LLC
    • Lou Hafer, Simon Fraser University,

Title: John Forrest-fest | COIN-OR 10th (Joint Cluster Computing)

  • Chair: Robin Lougee,Program Manager, COIN-OR, IBM TJ Watson Research Center, 1101 Kitchawan Road, Yorktown Heights NY 10598, United States of America, rlougee@…
  • Co-Chair: Alan King,IBM Research, PO Box 218, Yorktown Heights NY 10598, United States of America, kingaj@…
  • Here's the cluster as recorded in the INFORMS on-line system on Weds, Sept 8, 2010

Sunday Nov 07, 08:00 - 09:30 : Interfaces and Metaheuristics in COIN-OR Chair: Joao Goncalves,IBM Research, 1101 Kitchawan Rd, Yorktown Heights, United States of America, jpgoncal@…

  • Title: The Optimization Services Project
    • Presenting Author: Kipp Martin,University of Chicago, 5807 South Woodlawn, Chicago IL 60637, United States of America, kmartin@…
    • Abstract: In this talk we provide an update on recent developments in the COIN-OR project Optimization Services.

  • Title: Modeling Cone Optimization Problems with COIN OS
    • Presenting Author: Imre Polik,Lehigh University, Department of Industrial and Systems Engineering, 200 West Packer Avenue, Bethelehem PA 18015, United States of America, imre@…
    • Abstract: We present the extension of COIN OS to model cone optimization problems. We discuss the design principles of the extension. In the end we are able to model problems with matrices, thus extendeing the general modelling capabilities of COIN OS. This is joint work with Kipp Martin, Horand Gassmann and Jun Ma.

  • Title: Pulp a Linear Programming Toolkit for Python
    • Presenting Author: Stuart Mitchell,Research Fellow, University of Auckland, Private Bag 92019, Auckland 1142, New Zealand, s.mitchell@…
    • Abstract: PuLP https://www.coin-or.org/PuLP/ is a Toolkit for modelling and solving linear and mixed integer programs in Python. PuLP was created by J.S. Roy and released under a MIT licence. PuLP development focuses on ease of use and quick development. Dippy is a recent extension of PuLP to interface with DIP https://projects.coin-or.org/Dip a framework for implementing a variety of decomposition-based branch-and-bound algorithms for solving mixed integer linear programs.

  • Title: METSlib: A Metaheuristics Framework in Modern C++
    • Presenting Author: Mirko Maischberger,Università degli Studi di Firenze, Dipartimento di Sistemi e Informatica, Via S. Marta, 3, Firenze 50129, Italy, maischberger@…
    • Abstract: As framework for the rapid development of search strategies based on the exploration of a neighborhood METSlib offers the basic building blocks to assemble customized solvers using fully modular algorithms such as Tabu Search, SA and LS, combined in complex strategies. The user can implement a single model and neighborhood and easily try different algorithmic variations. We also present a Parallel Iterated TS solver for some extensions of the Vehicle Routing problem.

Sunday Nov 07, 11:00 - 12:30 : Open-Source Software Chair: Leo Lopes,Monash University Maths, Australia, leo.lopes@…

  • Title: On Generating Relevant Synthetic Instances for Practical Problems
    • Presenting Author: Leo Lopes,Monash University Maths, Australia, leo.lopes@…
    • Co-Author: Kate Smith-Miles,Professor, School of Mathematical Sciences, Monash University, Clayton, VIC 3800, Australia, kate.smith-miles@…
    • Abstract: The features of instances arising from practice are jointly distributed in complex ways that are hard for synthetic generators to replicate. Using the wrong instances to test solvers has led to deployment disasters. To address this, we present a methodology that helps improve instance generators and a statistic that summarizes the generator's ability to produce realistic and discriminating instances. The methodology is supported by an implementation in R available as a Coin-OR project.

  • Title: OpenSolver - An Open Source Solver-compatible CBC-based Excel Optimizer
    • Presenting Author: Andrew Mason,Dept of Engineering Science, University of Auckland, Private Bag 92019, Auckland, New Zealand, a.mason@…
    • Abstract: As part of a project on staff scheduling, we have developed OpenSolver, an open-source Excel addin that allows users to optimise their spreadsheet-based Solver models using the Coin-OR CBC integer programming solver. OpenSolver also provides tools for viewing an optimization model directly on a spreadsheet, and for quickly solving related models that differ only in the values of their right hand sides. OpenSolver can be downloaded from www.opensolver.org

  • Title: Rima: An Experiment in Composing Mathematical Models from Parts
    • Presenting Author: Geoff Leyland,Director, Incremental Limited, 17 Bridgens Ave, Te Atatu Peninsula, Auckland 0610, New Zealand, geoff_leyland@…
    • Abstract: Rima ( http://code.google.com/p/rima/ ) is a yet-another tool for formulating mathematical models implemented in a scripting language. Rima's goal is to make it easy to compose a model from parts, and to create generic parts. To this end, models are defined symbolically, problem data is late bound, models and data can be overloaded, and there are constructs for easy data transformation. Rima is implemented in Lua, binds to CLP, CBC and lpsolve, and is awaiting review to become part of COIN.

  • Title: Impact of Design Architecture and Developer Networks on Software Project Success
    • Presenting Author: Michelle Liu,Assistant Professor of IT, Marymount University, 2807 N. Glebe Rd., Arlington VA 22207, United States of America, xiangl2001@…
    • Abstract: In this study, the author seeks to understand the factors differentiating successful open source software projects from unsuccessful ones. Using data gathered from SourceForge.net, the author empirically tests several hypotheses about the impact of architecture and developer’s access to knowledge on a project’s success. Theoretical and practical implications are discussed, and several future research directions are outlined.

Sunday Nov 07, 13:30 - 15:00 : CEIR and Beyond Chair: John Tomlin,Principal Research Scientist, Yahoo! Research, 4301 Great America Parkway, Santa Clara CA 95054, United States of America, tomlin@…

  • Title: One LP Code a Minute
    • Presenting Author: John Forrest,IBM Research (Retd.), john.forrest@…
    • Abstract: A short, content free, look back at LP codes which I have used, abused and written. Some attempt may be made to put the more successful codes in the context of the available computer architectures, but more emphasis will be on the codes which nobody else has heard of and which I have tried to forget. For example, hands up all those who have heard of Alligator?

  • Title: From LP/90/94 to UMPIRE
    • Presenting Author: John Tomlin,Principal Research Scientist, Yahoo! Research, 4301 Great America Parkway, Santa Clara CA 95054, United States of America, tomlin@…
    • Abstract: This talk outlines the features of CEIR LP/90/94 that survived into the UMPIRE Mathematical Programming System, describes some that did not, and summarizes the new features that were introduced in this code.

  • Title: IBM's Optimization Subroutine Library - History and Innovations
    • Presenting Author: David Jensen,IBM Research, 1101 Kitchawan Road, Yorktown Heights, United States of America, davjen@…
    • Abstract: John Forrest was the principle author of IBM's Optimization Subroutine Library. It represented a significant departure from previous offerings for mathematical optimization. In this talk we will discuss the some of the innovations that made the product successful, focusing on the many contributions that John made to its design and implementation.

  • Title: Some OSL History
    • Presenting Author: J.P. Fasano,IBM Research Watson, 1101 Kitchawan Road, Yorktown Heights, United States of America, jpfasano@…
    • Abstract: This talk will provide some anecdotal stories and history on the early days of OSL development.

Sunday Nov 07, 16:30 - 18:00 : Forrest-fest | COIN-OR 10th Retrospective Chair: Robin Lougee,Program Manager, COIN-OR, IBM TJ Watson Research Center, 1101 Kitchawan Road, Yorktown Heights NY 10598, United States of America, rlougee@…

  • Title: Panel Discussion: Self-documenting Code and Other 'Tricks' of the Trade
    • Presenting Author: Robin Lougee,Program Manager, COIN-OR, IBM TJ Watson Research Center, 1101 Kitchawan Road, Yorktown Heights NY 10598, United States of America, rlougee@…
    • Co-Author: John Forrest,IBM Research (Retd.), john.forrest@…, John Tomlin,Principal Research Scientist, Yahoo! Research, 4301 Great America Parkway, Santa Clara CA 95054, United States of America, tomlin@…
    • Abstract: COIN-OR's success in its first decade is due in large part to the software contributions of John Forrest, particularly CLP and CBC. In this panel, John and others will share their perspectives on optimization software development – what we’ve learned and where we need to go. (Submit questions at info@…)

Monday Nov 08, 08:00 - 09:30 : Optimization Software for Problems Arising in Machine Learning Chair: Katya Scheinberg,Research Sceintist, Columbia University, Mudd bldg, 500 W 120th street, New York NY 10027, United States of America, ks79@…

  • Title: Fast Alternating Linearization Method for Sparse Inverse Covariance Selection
    • Presenting Author: Shiqian Ma,Columbia University, 500 W. 120TH ST, Mudd Blvd, RM 313, New York NY 10027, United States of America, sm2756@…
    • Co-Author: Donald Goldfarb,Professor, Columbia University, 500 W. 120TH ST, Mudd Building, RM 313, New York NY 10027, United States of America, goldfarb@…, Katya Scheinberg,Research Sceintist, Columbia University, Mudd bldg, 500 W 120th street, New York NY 10027, United States of America, ks79@…, Zaiwen Wen,NSF Math Institutes' postdoc, United States of America, zw2109@…
    • Abstract: Learning the structure of the genetic regulatory network is usually transformed to estimating a sparse inverse covariance (SICS) matrix from the sample data. We propose a novel first-order method based on alternating linearization technique for solving SICS. We establish iteration complexity bound for the proposed algorithm. Numerical experiments are conducted to demonstrate the practical efficiency of our algorithm.

  • Title: Efficient Algorithms for the Group Lasso
    • Presenting Author: Zhiwei (Tony) Qin,Columbia University, 313A, 500 West 120th Street, New York NY 10027-6699, United States of America, tonyqin@…
    • Co-Author: Donald Goldfarb,Professor, Columbia University, 500 W. 120TH ST, Mudd Building, RM 313, New York NY 10027, United States of America, goldfarb@…, Katya Scheinberg,Research Sceintist, Columbia University, Mudd bldg, 500 W 120th street, New York NY 10027, United States of America, ks79@…
    • Abstract: First we propose a general version of the block coordinate descent (BCD) algorithm to the Group Lasso with an efficient approach to perform subproblem optimization. As an alternative we propose an extension of ISTA/FISTA based on variable step-lengths as a simplified version of BCD. Both proposed algorithms show excellent performance over other state-of-the-art approaches to this problem. Additionally, we apply the algorithms to the Multiple Measurement Vector (MMV) recovery problem.

  • Title: Fast Alternate Linearization Methods for Robust Principal Component Analysis
    • Presenting Author: Donald Goldfarb,Professor, Columbia University, 500 W. 120TH ST, Mudd Building, RM 313, New York NY 10027, United States of America, goldfarb@…
    • Abstract: Robust principal component analysis (RPCA) problem can be solved as a convex optimization problem under fairly mild conditions. We present a fast alternating linearization method for doing this and an accelerated variant that obtains, in the latter case, an epsilon-optimal solution in O(1/sqrt(epsilon)) iterations. We apply our methods to solve applications of the RPCA model in surveillance video and video denoising that result in problems with up to 46 million variables.

  • Title: Fast First Order Method with Line Search
    • Presenting Author: Katya Scheinberg,Research Sceintist, Columbia University, Mudd bldg, 500 W 120th street, New York NY 10027, United States of America, ks79@…
    • Co-Author: Donald Goldfarb,Professor, Columbia University, 500 W. 120TH ST, Mudd Building, RM 313, New York NY 10027, United States of America, goldfarb@…
    • Abstract: We show a theoretically justified approach that allows us to use the line search in conjunction with an accelerated gradient method.

Monday Nov 08, 11:00 - 12:30 : Solving Nonconvex Optimization Problems with Couenne, an Open-Source Branch-and-cut Solver Chair: Pietro Belotti,Clemson University, Clemson SC 29634, United States of America, pbelott@…

  • Title: Rounding-based Heuristics for Nonconvex MINLPs
    • Presenting Author: Giacomo Nannicini,nannicin@…
    • Co-Author: Pietro Belotti,Clemson University, Clemson SC 29634, United States of America, pbelott@…
    • Abstract: We propose two primal heuristics for possibly nonconvex MINLPs, based on the idea of rounding the solution to a continuous NLP subject to linear constraints. We design a heuristic that tries to find a first feasible solution, and an improvement heuristic which starts at a feasible solution and tries to compute one with a better objective function value within a neighborhood of the starting point. Computational results show the effectiveness in practice of these simple ideas.

  • Title: Symmetry Breaking in Couenne
    • Presenting Author: Jim Ostrowski,jostrows@…
    • Co-Author: Pietro Belotti,Clemson University, Clemson SC 29634, United States of America, pbelott@…, Leo Liberti,Associate Professor, CNRS LIX, Ecole Polytechnique, LIX, Ecole Polytechnique,F-91128, Palaiseau, France, liberti@…
    • Abstract: In this talk we will discuss the recent addition of symmetry breaking methods to the MINLP solver Couenne. We will describe how Couenne computes the symmetry group of nonlinear programming problems and how Couenne uses orbital branching to exploit symmetry. Computational results using symmetric instances in the MINLPlib library will (hopefully) be provided.

  • Title: Lower and Upper Bounding Techniques in Couenne
    • Presenting Author: Pietro Belotti,Clemson University, Clemson SC 29634, United States of America, pbelott@…
    • Abstract: We discuss the recent advances in the cutting planes separators and in the heuristics available in Couenne, an Open-Source solver for Global Optimization that has been developed under the COIN-OR framework. In particular, we present a linearization technique for multilinear sets and present preliminary computational results that prove its usefulness in the context of bilinear and polynomial nonconvex optimization.

  • Title: Solving Linear Relaxations of Quadratically Constrained Quadratic Programs
    • Presenting Author: Andrea Qualizza,Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh PA 15213, United States of America, qualizza@…
    • Co-Author: Pietro Belotti,Clemson University, Clemson SC 29634, United States of America, pbelott@…, Francois Margot,Professor, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh PA 15213, United States of America, fmargot@…
    • Abstract: We discuss Linear Programming relaxations of Non-Convex Quadratically Constrained Quadratic Programs. We use a cutting plane method to approximate the well known PSD+RLT relaxations in the lifted space containing the Yij=xi*xj variables. We study linear inequalities arising from the outer-approximation of the cone of PSD matrices and sparsification of these inequalities derived from principal minors. Computational results based on instances from the literature will be presented.

Monday Nov 08, 13:30 - 15:00 : COIN-OR's Coopr Package: Highlights and Applications Chair: William Hart,Sandia National Laboratories, PO Box 5800, Albuquerque NM, United States of America, wehart@…

  • Title: Recent Developments in Pyomo
    • Presenting Author: William Hart,Sandia National Laboratories, PO Box 5800, Albuquerque NM, United States of America, wehart@…
    • Co-Author: Jean-Paul Watson,Sandia National Laboratories, jwatson@…
    • Abstract: The Pyomo modeling package can be used to formulate optimization models using the Python scripting language. Pyomo is being actively developed to support COIN-OR users in a variety of application areas. This talk will highlight recent developments in Pyomo, as well as new Coopr capabilities that can be leveraged to analyze Pyomo models.

  • Title: PySP: Modeling and Solving Stochastic Mixed-Integer Programs in Python
    • Presenting Author: Jean-Paul Watson,Sandia National Laboratories, jwatson@…
    • Co-Author: William Hart,Sandia National Laboratories, PO Box 5800, Albuquerque NM, United States of America, wehart@…, David Woodruff,Professor, University of California Davis, Davis CA, United States of America, dlwoodruff@…
    • Abstract: We describe PySP, an open-source extension of Pyomo - a Python-based modeling language for mathematical programming - that enables modeling and solution of stochastic mixed-integer programs. PySP contains a number of generic decomposition-based solution strategies made possible through Python language features such as introspection. We discuss the design and implementation of these generic strategies, in addition to computational results on standard stochastic benchmarks.

  • Title: Representation and Solution of Generalized Disjunctive Programming Problems using Pyomo
    • Presenting Author: Daniel Word,Texas A&M University, 3122 TAMU, College Station TX 77843, United States of America, dword@…
    • Co-Author: William Hart,Sandia National Laboratories, PO Box 5800, Albuquerque NM, United States of America, wehart@…, Carl Laird,Texas A&M University, 3122 TAMU, College Station TX 77843, United States of America, carl.laird@…, John Siirola,Sandia National Laboratories, P.O. Box 5800, Albuquerque NM, United States of America, jdsiiro@…
    • Abstract: Generalized Disjunctive Programming (GDP) is a common technique used in mathematical programming to describe discrete decisions in optimization problems, however, converting this representation to an MILP can be tedious and error-prone. We develop a plug-in for the python-based modeling language Pyomo that accepts GDP problems directly, and automatically performs the necessary big-M or convex-hull transformations. Example problems demonstrate the capabilities of this software.

  • Title: Global Optimization for Estimation of On/Off Seasonality in Infectious Disease Spread Using Pyomo
    • Presenting Author: Gabe Hackebeil,Texas A&M University, 3122 TAMU, College Station TX 77843, United States of America, gabehack@…
    • Co-Author: Carl Laird,Texas A&M University, 3122 TAMU, College Station TX 77843, United States of America, carl.laird@…
    • Abstract: Estimating an unknown On/Off seasonal input to infectious disease models is challenging because of the nonlinear equality constraints governing the dynamics and the need for discrete decisions to represent the On/Off behavior. Using the Pyomo framework, we can easily manage construction of linear underestimators and piecewise linear overestimators within a scripting language (python) that allows straightforward implementation of the global optimization algorithm.

Monday Nov 08, 16:30 - 18:00 : Recent Developments in COIN-OR Projects Chair: Laszlo Ladanyi,CPLEX R&D, IBM, P.O. Box 218, Yorktown Heights NY 10598, United States of America, ladanyi@…

  • Title: Coin Graph Classes: Creating Efficient and Extensible Graph Storage and Algorithms for C++
    • Presenting Author: Philip Walton,Engineering Manager, JBHunt Transport, Inc., 411 NW A St, Bentonville AR 72712, United States of America, hpwalton@…
    • Abstract: There are many factors to take into consideration when writing a general purpose graph library. There are theoretical concerns such as complexity of search and traversals, space complexity in terms of compactness of storage and general usability. This will cover those topics and how they are accounted for in the COIN-OR Graph Classes (CGC) as well as some more practical aspects of developing such a library in C++.

  • Title: PFunc: Modern Task Parallelism For High Performance Computing
    • Presenting Author: Anju Kambadur,IBM T J Watson Research Center, 1101 Kitchawan Road, Yorktown Heights NY 10598, United States of America, pkambadu@…
    • Abstract: Task parallelism is a powerful parallel programming model that allows parallelization of irregular algorithms. However, existing task parallel tools such as Cilk do not provide the flexibility required by expert programmers to exploit the power of the multi-core processors to their fullest. In this talk, we introduce PFunc, a novel library for expressing shared-memory task parallelism in C and C++ that has been designed to overcome the shortcomings of existing task parallel tools.

  • Title: Progress Update on COIN/DIP - Decomposition for Integer Programming
    • Presenting Author: Matthew Galati,Optimization Interface Lead, SAS Institute, 1400 Morris Drive, Suite 201, Chesterbrook PA 19087, United States of America, matthew.galati@…
    • Co-Author: Ted Ralphs,Associate Professor, Lehigh University, 200 West Packer Avenue, Bethlehem PA 18015, United States of America, ted@…
    • Abstract: COIN/DIP (Decomposition for Integer Programming) is an open-source extensible software framework for implementing decomposition-based bounding algorithms for use in solving large-scale discrete optimization problems. We will present a progress update on the development of the application MILPBlock as a fully automated (black-box) MILP solver using branch-price-and-cut.

Tuesday Nov 09, 08:00 - 09:30 : Modeling and Computation in Stochastic Programming Chair: Ellis Johnson,Professor, Georgia Tech, School of ISyE, Atlanta GA 30332-0205, United States of America, ellis.johnson@…

  • Title: Computational Study for Chance-constrained Stochastic Programs
    • Presenting Author: Feng Qiu,Student, Georgia Tech, ISyE, Atlanta GA 30332-0205, United States of America, fqiu@…
    • Co-Author: Shabbir Ahmed,Associate Professor, Georgia Institute of Tech, 765 Ferst Drive NW, Atlanta GA, United States of America, sahmed@…, Santanu Dey,Assistant Professor, Georgia Institute of Technology, 765 Ferst Drive NW, Atlanta GA 30332, United States of America, santanu.dey@…
    • Abstract: We propose a branch and cut method for solving linear programs with k-out-of-m constraints. These problems arise in sample average approximations of chance constrained stochastic programs. The method is based on exploiting star convexity structure and monotonicity of the problem in an integer programming formulation of the problem.

  • Title: A Stochastic Programming Approach to Air Traffic Flow Management
    • Presenting Author: Yu-Heng Chang,Georgia Institute of Technology, Industrial and Systems Engineering, Atlanta, United States of America, yhchang@…
    • Co-Author: John-Paul Clarke,Georgia Institute of Technology, School of AE, Atlanta, United States of America, johnpaul@…, Ellis Johnson,Professor, Georgia Tech, 765 Ferst Dr NW, GA-Tech ISyE, Atlanta GA 30332-0205, United States of America, ejohnson@…, Senay Solak,University of Massachusetts Amherst, Isenberg School of Management, Amherst MA, United States of America, solak@…
    • Abstract: We study the air traffic flow management problem in the presence of weather, which determines optimally how and when to send the aircraft towards an airspace sector and how aircraft react in response to the weather realizations. Two-stage and multistage stochastic programs are proposed, a rolling horizon method to approach the problems is presented, and Lagrangian relaxation and subgradient methods are suggested to justify the rolling horizon method.

  • Title: Optimization of R&D Project Portfolios under Endogenous Uncertainty
    • Presenting Author: Senay Solak,University of Massachusetts Amherst, Isenberg School of Management, Amherst MA, United States of America, solak@…
    • Co-Author: Earl Barnes,Georgia Tech, Atlanta GA, United States of America, ebarnes@…, John-Paul Clarke,Georgia Institute of Technology, School of AE, Atlanta, United States of America, johnpaul@…, Ellis Johnson,Professor, Georgia Tech, 765 Ferst Dr NW, GA-Tech ISyE, Atlanta GA 30332-0205, United States of America, ejohnson@…
    • Abstract: We consider the problem of optimizing a portfolio of R&D projects with several stochastic attributes. A formal description of the corresponding decision process is introduced, which leads to the modeling of the problem as a multistage stochastic integer program with decision dependent uncertainty realizations. Efficient solution procedures based on sample average approximation and Lagrangian decomposition are then developed to solve the resulting large scale problem.

Tuesday Nov 09, 11:00 - 12:30 : Decomposition in Linear and Stochastic Programming Chair: Alan King,IBM Research, PO Box 218, Yorktown Heights NY 10598, United States of America, kingaj@…

  • Title: Expected Utility Maximization in Large-scale Portfolio Optimization
    • Presenting Author: Gerd Infanger,Consulting Professor, Stanford University, Stanford CA 94305, United States of America, infanger@…
    • Abstract: Expected utility based portfolio optimization in conjunction with a factor model representation of asset returns is introduced and an equilibrium pricing model of asset returns based on an expected utility and factor based portfolio model calibrated to the market is presented. Numerical results showing significant advantages of expected utility and factor based portfolio optimization (large-scale, long- only, long-short) compared to mean-variance portfolio optimization are presented.

  • Title: Linearization Methods for Structural Estimation Problems
    • Presenting Author: John Birge,University of Chicago, Booth School of Business, Chicago IL 60637, United States of America, John.Birge@…
    • Abstract: Structural estimation problems attempt to determine the structural parameters of an underlying model based on observations of an agent's actions. The resulting mathematical structure poises challenges for computation, particularly in higher dimensions. This talk will discuss properties of the problem where linearization methods can be applied and alternative relaxations for more general problems that preserve convexity.

  • Title: Two-stage stochastic linear programs with Conditional Value at Risk (CVaR) constraints
    • Presenting Author: Dharmashankar Subramanian,Research Staff Member, IBM Research, 1101 Kitchawan Rd, Rte 134, Yorktown Heights NY 10598, United States of America, dharmash@…
    • Co-Author: Pu Huang,Research Staff Member, IBM Research, 1101 Kitchawan Rd, Rte 134, Yorktown Heights NY 10598, United States of America, pu.huang@…
    • Abstract: We consider a two stage stochastic programming problem with unresolved uncertainty that remains at the end of the second stage. We capture risk aversion by imposing a constraint on the two-stage composition of the CVaR risk measure. We present a decomposition algorithm for the above problem in a special case, where the induced risk-neutral version, i.e. the above formulation without the risk aversion (composed CVaR) constraint, can be tractably solved in its extensive (i.e. monolithic) form.

  • Title: Adventures in Stochastic Programming
    • Presenting Author: Alan King,IBM Research, PO Box 218, Yorktown Heights NY 10598, United States of America, kingaj@…
    • Co-Author: Gyana Parija,IBM India Research Lab, New Delhi, India, gyana.parija@…
    • Abstract: We discuss how nested parallel decomposition was implemented in OSL Stochastic Edition, and prospects for new adventures in COIN-OR.

Tuesday Nov 09, 13:30 - 15:00 : Solver Interface APIs I Chair: Matthew Saltzman,Clemson University, Dept. of Mathematical Sciences, Martin Hall, Box 340975, Clemson SC 29631, United States of America, mjs@…

  • Title: New CoinMP Release 2.0: A Simple C-API Windows DLL and Unix Solver Library (LP/MIP) based on COIN
    • Presenting Author: Bjarni Kristjansson,President, Maximal Software, Inc., 933 N. Kenmore St., Suite 218, Arlington VA 22201, United States of America, bjarni@…
    • Abstract: The COIN Open Source Initiative has become very popular in the recent years. To make life easier for users that simply want to solve models and not compile C++ applications, we have developed a standard C-API Windows DLL CoinMP.DLL that implements most of the functionality of CLP, CBC, and CGL. A Linux/Unix version using AutoMake is also available.

  • Title: GMO: GAMS' Next-Generation Model API
    • Presenting Author: Steven Dirkse,Director of Optimization, GAMS Development Corp., 1217 Potomac St. NW, Washington DC 20007, United States of America, sdirkse@…
    • Co-Author: Michael Bussieck,GAMS Development Corp., 1217 Potomac St. NW, Washington DC 20007, United States of America, mbussieck@…, Jan Jagla,GAMS Development Corp., 1217 Potomac St. NW, Washington DC 20007, United States of America, jjagla@…, Alex Meeraus,President, GAMS Development Corp., 1217 Potomac St. NW, Washington DC 20007, United States of America, ameeraus@…, Paul Vandereijk,GAMS Development Corp., 1217 Potomac St. NW, Washington DC 20007, United States of America, pvandereijk@…, Lutz Westermann,GAMS Development Corp., 1217 Potomac St. NW, Washington DC 20007, United States of America, lwestermann@…
    • Abstract: The GAMS model API plays a key role in linking solvers to GAMS. As solution technology & computing environments change, the model API must change with them. In this talk we describe the new API, its underlying philosophy, and the features required (both new and old) to create high-quality solver links on today's computers.

  • Title: Overview of and Update on ASL, the AMPL/Solver Interface Library
    • Presenting Author: David M. Gay,AMPL Optimization LLC, 900 Sierra Place SE, Albuquerque NM 87108-3379, United States of America, dmg@…
    • Abstract: The AMPL language processor communicates via files with solvers, including a ".nl" file of instance details. The AMPL/Solver interface library (ASL) has readers of .nl files that read this information and set up data structures for function evaluations carried out by ASL routines. It also provides a ".sol" file writer for returning solution details to the AMPL language processor. This talk provides an overview of the ASL, including some recent extensions, e.g., for stochastic programming.

Tuesday Nov 09, 16:30 - 18:00 : COIN-OR Under the Hood Chair: Kipp Martin,University of Chicago, 5807 South Woodlawn, Chicago IL 60637, United States of America, kmartin@…

  • Title: COIN-OR: Revving up the Engine
    • Presenting Author: Ted Ralphs,Associate Professor, Lehigh University, 200 West Packer Avenue, Bethlehem PA 18015, United States of America, ted@…
    • Abstract: In this talk, our goal will be to make some sense out of the bewildering array of projects in the COIN-OR repository and to show you how to get the COIN engine running smoothly within your applications.
  • Title: Build It Your Way: Building COIN with Visual Studio
    • Presenting Author: Lou Hafer, Simon Fraser University, Burnaby BC V5A 1S6, Canada, lou@…
    • Abstract: Plain old 32-bit lib won't do? This talk will discuss the possibilities and limitations for building COIN projects as libs, DLLs, and assemblies with Visual Studio.

  • Title: Getting Started with Python COIN-OR Software
    • Presenting Author: William Hart,Sandia National Laboratories, PO Box 5800, Albuquerque NM, United States of America, wehart@…
    • Co-Author: Stuart Mitchell,Research Fellow, University of Auckland, Private Bag 92019, Auckland 1142, New Zealand, s.mitchell@…
    • Abstract: COIN-OR has recently integrated several Python-based modeling packages: PuLP and Coopr. Installing and running Python software is a bit different than traditional software packages. This talk would discuss options for installation, integration with COIN-OR solvers and libraries, and future directions related to CoinAll integration.

  • Title: Coin Help
    • Presenting Author: Kipp Martin,University of Chicago, 5807 South Woodlawn, Chicago IL 60637, United States of America, kmartin@…
    • Abstract: In this talk we describe the CoinHelp project that is designed to make it easy for users to get up and running with COIN-OR.

Wednesday Nov 10, 08:00 - 09:30 : Solver APIs II Chair: Matthew Saltzman,Clemson University, Dept. of Mathematical Sciences, Martin Hall, Box 340975, Clemson SC 29631, United States of America, mjs@…

  • Title: The COIN-OR Open Solver Interface: A Status Report
    • Presenting Author: Matthew Saltzman,Clemson University, Dept. of Mathematical Sciences, Martin Hall, Box 340975, Clemson SC 29631, United States of America, mjs@…
    • Abstract: The Open Solver Interface (OSI) is the oldest multi-solver API in COIN-OR. We discuss its current status and plans for the future.

  • Title: A COIN OSI Solver Plugin for Microsoft Solver Foundation
    • Presenting Author: Lou Hafer, Simon Fraser University, Burnaby BC V5A 1S6, Canada, lou@…
    • Abstract: MsfOsi implements a solver plugin that allows Microsoft Solver Foundation to use COIN solvers through the OSI API. There are interesting differences in design philosophy. This talk will examine what it takes to bridge the gap.

  • Title: The Optimization Services Solver Interface
    • Presenting Author: Kipp Martin,University of Chicago, 5807 South Woodlawn, Chicago IL 60637, United States of America, kmartin@…
    • Abstract: In this talk we describe how to use the Optimization Services (OS) project to interface with COIN-OR solvers. The OS interface is quite flexible and allows the user to generate linear and nonlinear instances for solvers. In addition, there is an interface for solver options and solver results.

Wednesday Nov 10, 11:00 - 12:30 : Panel: COIN-OR Technology Forum Chair: Ted Ralphs,Associate Professor, Lehigh University, 200 West Packer Avenue, Bethlehem PA 18015, United States of America, ted@…

  • Title: Panel Discussion: COIN-OR Technology Forum
    • Presenting Author: Ted Ralphs,Associate Professor, Lehigh University, 200 West Packer Avenue, Bethlehem PA 18015, United States of America, ted@…
    • Co-Author: Lou Hafer,Simon Fraser University, Burnaby BC V5A 1S6, Canada, lou@…, William Hart,Sandia National Laboratories, PO Box 5800, Albuquerque NM, United States of America, wehart@…, Kipp Martin,University of Chicago, 5807 South Woodlawn, Chicago IL 60637, United States of America, kmartin@…
    • Abstract: Following up on last year's successful forum, this panel discussion will be an opportunity for users and developers of COIN-OR software to discuss recent and future developments within COIN-OR. If you want to get involved, provide feedback, or just learn about COIN-OR, please join us!

Last modified 7 years ago Last modified on Sep 28, 2010 7:54:16 PM

Attachments (4)

Download all attachments as: .zip