 |
|
|
|
|
|
|
|
|
|
 |
||||
|
|
|||
|
||||
|
||||
|
|
|
Presentations (chronological order) Horand Gassman (presenter), Jun Ma, Kipp Martin, Recent Development in Optimization Services (OS), (12/14/2016) pdf Recent Development in Optimization Services (OS) Horand Gassman (presenter), Jun Ma, Kipp Martin, Imre Polik, Extensions to the OSiL schema: Recent developments in Optimization Services (OS), (07/11/2016) pdf Optimization Services (OS) is an open-source project under COIN-OR and provides infrastructure for the solution of optimization problems over the internet. This includes a number of XML schemas for the transmission of problem instances, options and solutions as well as related information; and interfaces to a number of open-source and commercial solvers. This talk presents recent developments, such as support for the description of matrices within the OSiL format, an interface to the open-source SDP solver CSDP, and schema elements to describe real-time data elements, problem modifications, disjunctions, and stochastic information. Horand Gassman (presenter), Jun Ma, Kipp Martin, Imre Polik, Extensions to the OSiL schema: Matrix and cone programming, (07/17/2014) pdf ppt A lot of attention has been given recently to cone programming and matrix programming, using, for instance, relaxations of hard mixed integer programs using variables whose values are required to form symmetric positive semidefinite matrices or satisfy similar cone constraints. This talk presents efforts to facilitate the formulation of such problems within the OSiL framework, an XML schema used to allow a unified representation format for a large variety of mathematical optimization problem instances. OSiL is part of the OS project, an open source project under the COIN-OR umbrella. Horand Gassman (presenter), Jun Ma, Kipp Martin, Optimization Services: Communicating Solver Options and Solver Results, (11/15/2011) pdf Optimization Services (OS) is a web-aware framework linking algebraic modeling languages and solvers. Communication takes place in standardized XML formats. After a brief overview of the system design I will describe the option and result formats in more detail. Unlike the instance description, there are choices here to allow for both flexibility to cater to the varying design choices of solver developers, and rigidity to allow one AML to talk to multiple solvers without changing the interface. Mike Steglich, CMPL (Coliop/COIN Mathematical Programming Language), (11/15/2011) link CMPL is a mathematical programming language and a system for modelling, solving and analysing linear programming (LP) problems and mixed integer programming (MIP) problems. Lou Hafer, Mattew Saltzman (presenter), The New OSI: Lighter, Thinner, Stronger, More Dynamic, (11/15/2011) link Over the 10 years of its existence, the original OSI interface has grown--unlike most of its users--from 'just a little pudgy' to 'seriously overweight'. OSI2 is a return to something a bit more lean and mean, with outsourcing. In this talk we will discuss the implementation of an interface layer whose primary function is to act as a broker, managing modules provided by underlying solvers to implement defined APIs. Lou Hafer, William Hart, Kipp Martin, Ted Ralphs, Matthew Saltman , Panel Discussion: COIN-OR Technology Forum, (11/15/2011) pdf link 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! Jeffrey Camm, Saravanan Kuppusamy, Michael Magazine, Kipp Martin (presenter), Kipp Martin, The Bearcats Transportation System: A New Vehicle Routing Formulation (11/14/2011) pdf We consider the problem of constructing optimal bus routes. The objective is to minimize the distance traveled weighted by the number of riders on the route. The resulting model is a non-convex integer program that is extremely difficult to optimize. We show how to reformulate this problem as a linear integer program in an extended variable space. The resulting formulation is very large and we describe a column generation algorithm based on COIN-OR software. Horand Gassman (presenter), Jun Ma, Kipp Martin, Optimization Services: Communicating Solver Options and Solver Results (05/03/2011) pdf Optimization Services (OS) is a framework that links algebraic modelling languages (AMLs) and solvers, developed as an open source project under the umbrella of COIN-OR. OS supports a variety of linear and nonlinear solvers, both open source and commercial. OS is web-aware, allowing for seemless integration of distributed computing environments and even cloud computing. Communication between algebraic modelling languages and solvers takes place through strings and files in standardized XML formats. I will give a brief overview of the system design and will then concentrate on describing the option and result formats in more detail. Unlike the instance description, there are challenges here to allow for both flexibility to cater to the varying design choices of solver developers, and rigidity to allow one AML to talk to multiple solvers without changing the interface. Horand Gassman , Jun Ma, Kipp Martin (presenter), The Optimization Services Solver Interface (11/10/2010) pdf 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. Kipp Martin (presenter), Coin Easy (11/09/2010) pdf In this talk we describe the CoinEasy project that is designed to make it easy for users to get up and running with COIN-OR. Horand Gassman , Jun Ma, Kipp Martin (presenter), The Optimization Services Project (11/07/2010) pdf In this talk we provide an update on recent developments in the COIN-OR project Optimization Services. Horand Gassman , Jun Ma, Imre Pólik (presenter), Kipp Martin, Modeling Cone Optimization Problems with COIN OS (11/07/2010) 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. Horand Gassman , Jun Ma, Imre Pólik (presenter), Kipp Martin, Modeling Cone Optimization Problems with COIN OS (10/12/2009) pdf We present a new modelling extension of the popular COIN OS, to model cone optimization problems. Not only this extension is able to represent a wide range of optimization problems, but it can also capture the special structure of them. This is crucial, since it enables the solvers to exploit these special structures to speed up the computation or to improve accuracy. Saravanan Kuppusamy (presenter), Michael Magazine, Kipp Martin, Transportation System Design Using COIN-OR (10/12/2009) pdf The Bearcat Transportation System of the University of Cincinnati is modeled and solved as a multiple vehicle routing problem. The Student population and travel time data coupled with client driven constraints are used to design the transportation system. The COIN-OR solver CBC is used in a GAMS environment to solve the problem. Results indicate the solutions are effective and implementable. Craig Froehle, Michael Magazine, Kipp Martin (presenter), Cary Wise, Scheduling Doctors to Clinical and Surgical Time Slots: A Column Generation Approach 10/12/2009 pdf We consider the problem of scheduling doctors to clinical and surgical time slots over a five-week horizon. This leads to an integer program with millions of variables. We discuss the solution of this problem using column generation with COIN-OR branch cut and price (BCP). Horand Gassmann (Presenter), Jun Ma, and Kipp Martin, Recent Changes to the Optimization Services Project to Support Stochastic Programming (06/14/2009) ppt pdf Optimization Services (OS) is a COIN-OR project to support the modelling and solution of a wide range of optimization problems both via local servers and over a network. This talk describes recent additions to the system, particularly as they relate to supporting stochastic programming and the representation of uncertainty. Kipp Martin (Presenter), Horand Gassmann and Jun Ma, Optimization Services, Web Services, and Excel (01/12/2009) pdf Optimization Services (OS) is a framework for representing optimization instances, results, solver options, and communication between clients and solvers in a distributed environment using Web Services. In this talk we describe how we have incorporated the OS project within Microsoft Excel. We provide a user-defined class that can be used to build mixed integer linear programs inside Excel using Visual Basic for Excel. We also provide a class that allows the user to call COIN-OR solvers remotely from Excel using Web Services. Diego Klabjan (Presenter), Robert Fourer and Jun Ma, Algebraic Modeling in a Deductive Database Language (01/12/2009) ppt pdf Datalog is a deductive language tailored for easy database access. We present an algebraic modeling language in Datalog for mixed-integer linear optimization models. Using this language, data can be easily queried from a database and combined with models to produce problem instances, providing an advantage over conventional optimization modeling languages that rely on reading data by use of builtin tools or importing data from external sources via standard files. The declarative nature of Datalog permits the underlying syntax to be kept reasonably intuitive and simple. Molham Aref, David Zook and Emir Pasalic (Presenter), Using Optimization Services in Datalog (01/12/2009) ppt pdf Optimization Services is a recent web-based standard for optimization. Among other features, they allow flexible use of optimization solvers as plug-ins. LogicBlox is a commercial enterprise-level modeling and optimization software platform based on the Datalog programming language, with sophisticated back-and front-end interfacing. We discuss Datalog and the LogicBlox computing framework in the context of optimization services. H. Gassmann (presenter), J. Ma, K. Martin, Recent Developments in OSiL/SE (10/13/2008) ppt pdf OSiL is an XML schema for the formulation of a large class of mathematical programming problems. This talk concerns recent advances in the description of stochastic programs. Also described are a parser, internal data objects and a rudimentary implementation of a decomposition algorithm. B. Brad (presenter), Calculating Spare Hessian using Algorithmic Differentiation (10/13/2008) The CppAD package can use the fact that a Hessian is sparse to reduce the work necessary to calculate the Hessian. We will review how multiple columns of a sparse Hessian can be computed with the same work as one column. CppAD uses the greedy distance two graph coloring algorithm to group columns that can be computed together. We will review this algorithm and some speed tests of the sparse Hessian calculation. Finally we will suggest some ways in which this calculation can be improved. B. Bell, R. Lougee-Heimer, K. Martin (presenter), COIN-OR Vendor Workshop (10/11/2008) pdf R. Fourer (presenter), J. Ma, K. Martin, The Optimization Services Project on COIN-OR: Progress and Plans, IFORS Triennial Conference, Sandton, South Africa (07/14/2008) ppt pdf We describe Optimization Services (OS), a unified framework for a new generation of Internet optimization systems, which has advantages for both developers and users of optimization software. OS incorporates XML-based standards for representing and communicating optimization problems, so that each solver needs only one interface. Components of OS include an XML-based representation for problem instances and a corresponding inmemory representation. An open source library, hosted on COIN-OR, provides utilities for reading and writing these representations and for converting between them. B. Brad (presenter), C++ Algorithmic Differentiation by Operator Overloading, INFORMS Annual Conference, Seattle, Washington, USA 11/04/2007 web Reverse mode algorithmic differentiation calculates a gradient using a small multiple of the number of operations required to compute an objective. This requires storing the operations and values corresponding to the objective. As memory gets larger, reverse mode can be efficiently applied in more cases. We present a C++ templated objective and use it with float, double, and the AD types defined by ADOLC, CppAD, and FADBAD. The gradient computation time using these AD packages is compared. R. Fourer, J. Ma (presenter), K. Martin, Setting Up and Hosting Your Solver as Web Services via Optimization Services (OS), INFORMS Annual Conference, Seattle, Washington, USA 11/04/2007 ppt pdf Optimization Services (OS) is a unified framework for new generation distributed optimization systems. We provide the open source OS library which is based on a set of Optimization Services Protocols (OSP). The library supports the solution of a wide variety of optimization problem types in a loosely coupled distributed environment. We also provide OS server software for users to host their own Optimization Services. In particular, we demo the OS integration in Apache's Tomcat Web server. R. Fourer (presenter), J. Ma, K. Martin, Hooking Optimization Services to Modeling Languages and Solvers, INFORMS Annual Conference, Seattle, Washington, USA 11/04/2007 pdf1 pdf2 The OS framework is designed to avoid duplication of effort by means of a new standard for representing optimization problem instances. Once adopted, this standard will require only one interface to be written for each modeling system and each solver. At the heart of our proposal are OsiL, an XML-based representation for optimization problem instances, and OSInstance, a corresponding in-memory representation, together with an object-oriented library of methods for working with these forms. R. Fourer, J. Ma, K. Martin (presenter), Optimization Services and Nonlinear Optimization, INFORMS Annual Conference, Seattle, Washington, USA 11/04/2007 pdf We describe how to use Optimization Services (OS) with third party nonlinear solvers. The OS library contains an OS instance class that provides a flexible API for solvers. The OSInstance class has a method to generate instance representations in postfix format. This is illustrated with the LINDO solver. There are also calculate methods that use the CppAD package to provide gradient and Hessian values using algorithmic differentiation. This is illustrated with Knitro and Ipopt. R. Fourer, J. Ma, K. Martin (presenter), The COIN-OR Optimization Services Project, INFORMS Annual Conference, Seattle, Washington, USA 11/04/2007 pdf Optimization Services (OS) provides a set of standards for representing optimization instances, results, solver options, and communication between clients and solvers in a distributed environment. We provide source code for libraries that implement OS standards. The project includes an executable, OSSolverService, for communicating instances (OSiL XML format, AMPL nl format, MPS format) to solvers. Software is provided for communication with Cbc, Clp, Cplex, Glpk, Ipopt, Knitro, and LINDO. R. Fourer (presenter), How to Publish Your Software on COIN-OR, INFORMS Annual Conference, Seattle, Washington, USA 11/04/2007 web COIN-OR offers an environment tailor-made for disseminating and maintaining OR software. If you have created a software package or library that you are willing and able to "open up" for use by the OR community, you may want to consider adding it to the over two dozen projects currently supported at COIN-OR. This presentation will detail the steps involved. G. Gassman (presenter), R. Fourer, J. Ma, K. Martin, OSiL: An XML-based schema for stochastic programs, 11th Conference on Stochastic Programming (SPXI), Vienna, Austria, 08/30/2007 ppt pdf R. Fourer (presenter), J. Ma, K. Martin, An Open Interface for Hooking Solvers to Modeling Systems, INFORMS International, Puerto Rico, 07/10/2007 ppt pdf We describe a new open standard for representing optimization problem instances. At the heart of our proposal are OSiL, an XML-based representation for instances, and OSInstance, a corresponding in-memory representation. An open-source library, maintained as a COIN-OR project, provides application programming interfaces for reading and writing OSiL and OSInstance and for converting between them. Related representations handle the passing of options to solvers and the retrieval of results. R. Fourer (presenter), How to Publish Your Software on COIN-OR, INFORMS International, Puerto Rico, 07/10/2007 ppt pdf COIN-OR offers an environment tailor-made for disseminating and maintaining OR software. If you have created a software package or library that you are willing and able to "open up" for use by the OR community, you may want to consider adding it to the over two dozen projects currently supported at COIN-OR. This presentation will detail the steps involved. H. Sheng (presenter), J. Ma, Generalized Branching Hyperplane Methods and Distributed Computing Technologies, 07/09/2007 pdf We studied the generalized branching hyperplane methods and distributed computing technologies for mixed integer nonlinear programming research. R. Fourer (presenter), The NEOS Benchmarking Service, INFORMS International, Puerto Rico, 07/09/2007 ppt pdf One of the "solvers" on NEOS 5 is actually a benchmarking service that runs a submitted problem on a selection of solvers on the same machine. In addition to output listings from the solvers, the service optionally returns an independent assessment of the quality of the results. It is particularly useful in choosing a solver for a particular application. R. Fourer (presenter), J. Ma, K. Martin, The Optimization Services Project on COIN-OR, INFORMS Computing Society Conference, Coral Gables, FL, USA (01/03/2007) pdf Optimization Services (OS) is a unified framework for a new generation of distributed optimization systems,mainly for optimization over the Internet. It incorporates XML-based standards for representation and communication of optimization-related data between OS-compatible software components. We describe the open-source OS project and its implications for both developers and users of optimization software. R. Fourer, J. Ma (presenter), K. Martin, An Enterprise Computational System Built on the Optimization Services (OS) Framework and Standards, INFORMS Annual Conference, Pittsburgh, Pennsylvania, USA (11/08/2006) ppt pdf Optimization Services (OS) is a unified framework and a set of open source standards (under the CPL license) for building Comutational and Optimization software both as standalone and over distributed systems. In this talk we present an professional implementation and deployment of large scale enterprise computational system that is based on the OS standards and fit for scheduling computational jobs over both Internet and Intranet. R. Fourer, J. Ma (presenter), K. Martin, Using the Optimization Services hookup Language (OShL) to Invoke Remote Optimization Services, INFORMS Annual Conference, Pittsburgh, Pennsylvania, USA (11/08/2006) ppt pdf We describe an open-source, platform and language independent API that implements OShL, a universal framework for hosting and invoking optimization software as services. OShL makes use of Web service standards to facilitate synchronous ("solve") and asynchronous ("send") optimization via the Internet. OShL has built-in features to maintain states ("getJobID") between subsequent service invocations, and to remotely "knock", "kill", and "retrieve" optimization jobs from the services. H. Sheng (presenter), J. Ma, S. Mehrotra, Impact Solver for Optimization Services, INFORMS Annual Conference, Pittsburgh, Pennsylvania, USA (11/08/2006) ppt pdf We improve our IMPACT solver for broader range of optimization problems. New algorithms are being developed and tested for large scale mixed nonlinear integer problems. Combining with Optimization Services(OS) protocols, we plan to provide a robust solver service for applications of mathematical programming. J. Ma (presenter), S. Mehrotra, H. Sheng, On Implementing a Parallel Integer Solver Using Optimization Services, INFORMS Annual Conference, Pittsburgh, Pennsylvania, USA (11/08/2006) ppt pdf We will present a framework for implementing a parallel integer solver using optimization services and standard instances. Depending on the progress computational results will be presented describing the viability of this approach for implementing parallel solvers for solving difficult integer programs within the IMPACT solver package. R. Fourer, J. Ma, K. Martin (presenter), A Result Language (OSrL) and Solver Option Language (OSoL) for Distributed Optimization, INFORMS Annual Conference, Pittsburgh, Pennsylvania, USA (11/08/2006) pdf We present an XML-based result language (OSrL) and an XML-based solver option language (OSoL) for distributed optimization systems. We also describe associated in-memory objectes OSResult and OSOption that are used in a library to support client-solver communication in a system based on Web Services. R. Fourer (presenter), J. Ma, K. Martin, Extensions to an Optimization Services Instance Language, INFORMS Annual Conference, Pittsburgh, Pennsylvania, USA (11/08/2006) pdf Optimization problems of interest today go beyond the traditional linear, integer, quadratic, and smooth nonlinear types. A language for problem instances must be extended accordingly. This presentation describes prospective extensions to OSiL, our proposed language standard, in such areas as combinatorial optimization and constraint programming, stochastic programming, and semidefinite and cone programming. R. Fourer (presenter), J. Ma, K. Martin, The OSInstance Application Programming Interface for Optimization Problem Instances, INFORMS Annual Conference, Pittsburgh, Pennsylvania, USA (11/08/2006) pdf As a complement to our proposal for OSiL, a new XML-based representation for optimization problem instances, we propose OSInstance, an in-memory representation. An open-source library provides application programming interfaces for reading and writing OSiL and OSInstance and for converting between them. This arrangement allows the two standards to be defined by a single design, yet offers diverse “get” methods for extracting data from OSInstance objects in a variety of convenient ways. J.P. Fasano, K. Martin (presenter), T. Ralphs, Source Code Not Required: Using the COIN Binaries, INFORMS Annual Conference, Pittsburgh, Pennsylvania, USA (11/07/2006) pdf COIN-OR provides open-source software for the operations research community. However, many users may wish to take advantage of COIN-OR without downloading and compiling source code. There is a COIN-OR project that provides compiled libraries and executables for several key COIN-OR projects. These binaries are available for the Windows, Linux, and Mac OS X platforms. We describe the available binaries and illustrate how users can link to them and use them for solving optimization problems. G. Gassmann (presenter), Applied Stochastic Programming - Models and Computation, INFORMS Annual Conference, Pittsburgh, Pennsylvania, USA (11/07/2006) ppt pdf This talk gives a review of fifty years of stochastic programming, concentrating especially on applications, models and computation. Z. Li (presenter), S. Mehrotra, Extending An Interior-Point Algorithm with Higher Order Corrections to Semidefinite Programming, INFORMS Annual Conference, Pittsburgh, Pennsylvania, USA (11/07/2006) pdf We extend a primal-dual infeasible interior-point algorithm with higher order corrections for linear programming (LP) to Semidefinite Programming (SDP). The resulted algorithm has the capbility to combine higher order corrections generated by different strategies. The search direction of the algorithm can be obtained by solving a small SDP program. We give convergence conditions for the algorithm. R. Fourer (presenter), J. Ma, K. Martin, J. Moré, T. Munson, D. Orban, J. Sarich, Optimization Via the Internet: NEOS 5 and Beyond, 19th International Symposium on Mathematical Programming, Rio de Janeiro, Brazil (08/03/2006) pdf The NEOS Server provides free net access to nearly 50 "solvers" of optimization problems via a variety of interfaces and languages. It handles 3000-6000 requests a week and features a distributed design that scales readily. This survey emphasizes ideas behind the latest, completely rewritten version, which has an application programming interface that makes it callable from other programs. The presentation also offers a look at upcoming developments: automated problem analysis, XML-based standards for optimization problem instances, and an Optimization Services framework based on new web-service standards. J. Ma (presenter I), R. Fourer (presenter II, K. Martin (presenter III), AN OPEN INTERFACE FOR HOOKING SOLVERS TO MODELING SYSTEMS, DIMACS Workshop on COIN-OR, Rutgers University, Piscataway, New Jersey, USA (07/19/2006) Part I: Optimization Services (presented by J. Ma) ppt pdf Part II: An Open Interface for Hooking Solvers to Modeling Systems (presented by R. Fourer) pdf Part III: An Open Interface for Hooking Solvers to Modeling Systems (presented by K. Martin) pdf Every optimization modeling system has developed its own way of representing problem instances. Hence each solver package for optimization problems must have a separate "driver" for every modeling system that it supports. We describe a new open standard that, once adopted by modeling systems, will require each solver to have only one driver. The components of our proposal include OsiL, a new XML-based representation for optimization problem instances, and OSInstance, a corresponding in-memory representation. An open source library provides application programming interfaces for reading and writing OSiL and OSInstance and for converting between them. Related languages and libraries handle the passing of options to solvers and the retrieval of results, as well as protocols for facilitating solver registration, client and scheduler discovery, and communication between the various components of the optimization process. This session will emphasize the practical aspects of using our tools to hook solvers to modeling systems, focusing on our overall Optimization Services framework (Ma), the modeling system interface to our library (Fourer), and the solver interface to our library (Martin). M. Saltzman (presenter), The COIN-OR Open Solver Interface, DIMACS Workshop on COIN-OR, Rutgers University, Piscataway, New Jersey, USA (07/17/2006) pdf The COIN-OR Open Solver Interface provides C++ classes for accessing a variety of LP and MIP solvers using a single API. Calls are provided to read models from files or construct them in memory, modify them, invoke solvers, recover solution information, and write problems to files. We will describe features of the API and provide examples of its use. The long-term objective of the OSI project is to provide a portable, extensible prototype standard API for solvers for a variety of optimization problems. JP. Fasano (presenter), Getting and Building COIN-OR, DIMACS Workshop on COIN-OR, Rutgers University, Piscataway, New Jersey, USA (07/17/2006) ppt pdf In this talk we will discuss how to obtain the source code and build the modules available on the COIN-OR web site, www.coin-or.org. The goal of this session is to provide a hands-on opportunity for all attendees bringing wireless-enabled laptops to download and build the COIN-OR Linear Program Solver (CLP) and the COIN-OR Branch and Cut Solver (CBC). Because of the wide variety of platforms expected at the workshop, attendees are encouraged to try to download and build their favorite projects prior to the workshop. For assistance with individual issues or to suggest a topic for this talk, please contact the speaker prior to the workshop at jpfasano@us.ibm.com R. Fourer (presenter), J. Ma, K. Martin, OSiL: An Open Standard for Expressing and Using Optimization Problem Instances, EURO XXI —21st European Conference on Operational Research, Reykjavik, Iceland (07/05/2006) pdf Distributed modeling environments necessitate an open standard for exchanging optimization problem instances. For this purpose we present OSiL, an XML-based instance representation for large-scale linear and nonlinear optimization. OSiL uses the object-oriented features of XML to efficiently represent nonlinear expressions. Its schema maps directly to an in-memory representation that provides a robust application programming interface, facilitates reading and writing a range of data formats, and makes the nonlinear expression tree readily available for function and derivative evaluations. R. Fourer, G. Gassmann (presenter), J. Ma, K. Martin, An XML-based schema for stochastic programs, EURO XXI —21st European Conference on Operational Research, Reykjavik, Iceland (07/05/2006) ppt pdf This talk concerns the latest developments in an XML-based schema to describe stochastic programming problems. It is part of an on-going effort towards a unified problem description for all mathematical programming problems. System capabilities allow the formulation of recourse problems with and without deterministic problem dimensions, chance-constrained problems, alternate measures of risk, robust optimization problems, all subject to a number of built-in or user-specified probability distributions. The talk will describe the schema using illustrative examples. R. Fourer (presenter), J. Ma, K. Martin, J. Moré, T. Munson, D. Orban, J. Sarich, Optimization Via the Internet: NEOS 5 and Beyond, INFORMS International, Hong Kong, China (06/27/2006) pdf The NEOS Server provides free net access to nearly 50 "solvers" of optimization problems via a variety of interfaces and languages. It handles 3000-6000 requests a week and features a distributed design that scales readily. This survey emphasizes ideas behind the latest, completely rewritten version, and offers a look at upcoming developments: automated problem analysis, modern forms for optimization problem instances, and an Optimization Services framework based on new web-service standards. G. Gassmann (presenter), R. Fourer, J. Ma, K. Martin, An XML-based Schema for Stochastic Programs, APMOD 2006 (Applied Mathematical Programming and Modeling), Madrid, Spain (06/19/2006) ppt pdf This talk concerns the latest developments in an XML-based schema to describe stochastic programming problems. It is part of an on-going effort towards a unified problem description for all mathematical programming problems. System capabilities allow the formulation of recourse problems with and without deterministic problem dimensions, chance-constrained problems, alternate measures of risk, robust optimization problems, all subject to a number of built-in or user-specified probability distributions. The talk will describe the schema using illustrative examples. The NEOS Server provides free R. Fourer, J. Ma, K. Martin (presenter), Optimization Services: A Framework for Distributed Optimization, University of Cincinnati, in Cincinnati, Ohio, USA (06/02/2006) pdf Software-as-a-service is radically changing the traditional software model. Companies
are experiencing tremendous savings by using enterprise application software on an
as-needed basis rather than purchasing and installing the software. There is no reason why G. Gassmann (presenter), An introduction to stochastic programming, Industrial Engineering and Management Sciences, Northwestern University, Evanston, Illinois, USA (05/30/2006) ppt pdf 2005 marked the 50th anniversary of stochastic programming. In that year two articles were published (by G.B. Dantzig in Management Science and E.M.L. Beale in the Journal of the Royal Statistical Society) that formulated the first stochastic linear programs and gave algorithms for their solution. I will give an introduction to the field, using different problem formulations, and I will describe several general classes of algorithms. My particular interest lies in representations of stochastic programs, and I will describe different approaches, including the SMPS format, algebraic modeling languages, internal data structures, as well as a new approach using XML. J. Ma (presenter), Optimization Services(OS), IEMS Advisory Board Meeting, Industrial Engineering and Management Sciences, Northwestern University, Evanston, IL, USA (03/06/2006) ppt pdf From a high level business perspective, we present a general optimization system design introduced under our new concept of Optimization Services (OS) along with its Optimization Services Protocol (OSP). Optimization Services is intended to be a unified framework for the next generation distributed optimization systems, mainly optimization over the Internet. Thus Optimization Services can be regarded as the Operations Research Internet. The corresponding Optimization Services Protocol is intended to be a set of industrial standards. J. Ma, K. Martin (presenter), Optimization Services Modeling Language (OSmL), INFORMS Annual Conference, San Francisco, California, USA (11/15/2005) pdf We describe Optimization Services modeling Language (OSmL), an open source algebraic modeling language for mathematical optimization. OSmL is designed to convert raw data in XML format into problem instances that conform to the Optimization Services instance language (OSiL) standard. There is support for sets, indexing, and loops. An optimization model represented in OSiL can be solved with any solver that is Optimization Services compatible. R. Fourer, J. Ma, K. Martin (presenter), Optimization Services (OS) Framework, INFORMS Annual Conference, San Francisco, California, USA (11/15/2005) pdf We describe Optimization Services modeling Language (OSmL), an open source algebraic modeling language for mathematical optimization. OSmL is designed to convert raw data in XML format into problem instances that conform to the Optimization Services instance language (OSiL) standard. There is support for sets, indexing, and loops. An optimization model represented in OSiL can be solved with any solver that is Optimization Services compatible. R. Fourer, J. Ma (presenter), K. Martin, Optimization Services OS Server and OS Libraries, INFORMS Annual Conference, San Francisco, California, USA (11/15/2005) ppt pdf The OS library is an open source library based on a set of Optimization Services Protocols (OSP). The library supports the solution of a wide variety of optimization problem types in a loosely coupled distributed environment. Besides the OS library, we provide OS server software for users to host their own Optimization Services. We also mention a general-purpose Optimization Services modeling Language (OSmL). J. Ma, S. Mehrotra, H. Sheng (presenter), IMPACT Solver for Optimization Services, INFORMS Annual Conference, San Francisco, California, USA (11/15/2005) ppt pdf IMPACT is a new Integrated Math Programming Advanced Computational Tool. It is the first optimization solver that natively supports the Optimization Services (OS) protocols. It integrates state-of-the-art optimization algorithms with the OS framework to provide a robust solver service. R. Fourer (presenter), D. Orban, A Problem Instance Analyzer for Optimization Services, INFORMS Annual Conference, San Francisco, California, USA (11/15/2005) pdf We describe new developments in the design and testing of Dr. AMPL, a collection of utilities for determining properties of optimization problem instances generated from the AMPL modeling language. Dr. AMPL's problem analyzer checks properties ranging from size and sparsity to linearity and convexity, then compares the results to a database of solver characteristics to produce a list of recommended solvers. This information can then be fed to optimization services such as NEOS and OSxL. R. Fourer, G. Gassmann, B. Kristjansson, K. Martin, T. Ralphs, M. Saltzman, Panel Discussion: Standards for Optimization Problem Instances, INFORMS Annual Conference, San Francisco, California, USA (11/14/2005) Speakers from this and the preceding session will discuss standards issues that have been (or should have been) raised by the presentations., R. Fourer, J. Ma, K. Martin (presenter), M. Saltzman, Model Representation and an Open Solver Interface, INFORMS Annual Conference, San Francisco, California, USA (11/14/2005) pdf In order to solve optimization models it is necessary to put them in a solver-compatible format. This is a problem due to the large number of solvers. We present a C++ open-source solver interface (or API) that can be used to communicate model instances to a solver. R. Fourer, J. Ma (presenter), K. Martin, Optimization Services Instance Language (OSiL), INFORMS Annual Conference, San Francisco, California, USA (11/14/2005) ppt pdf In the area of mathematical optimization, it is becoming increasingly common to separate modeling languages from optimization solvers. This makes it critical to have an open standard for exchanging model instances. We present OSiL (Optimization Services Instance Language), an XML Schema for representing optimization instances. We also describe open-source Java libraries that simplify the exchange of problem-instance and solution information between modeling languages and solvers. R. Fourer, G. Gassmann (presenter), J. Ma, K. Martin, OSiL - Stochastic Extensions, INFORMS Annual Conference, San Francisco, California, USA (11/14/2005) ppt pdf This talk presents extensions to an XML schema to allow instance representations of a wide variety of stochastic programs. Included are multistage recourse problems with discrete and continuous distributions, different risk measures (such as VaR and CVaR), robust optimization models, etc. Also included are various means of modelling stochastic processes, including ARMA(p,q) processes. S. Mehrotra (presenter) , Z. Li, H. Sheng, Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming, INFORMS Annual Conference, San Francisco, California, USA (11/13/2005) ppt pdf We will present our computational experience with branching on hyperplane algorithms for mixed integer programming. Results will be discussed based on the performance of basis reduction algorithms and implementation strategies for different formulations of this approach. Results based on an implementation in a solver system "IMPACT" will be presented on difficult knapsack and market split problems; and depending on the progress, on sparse mixed integer programming problems. D. Elkins, R. Fourer, M. Grant, D. Heltne, R. Lougee-Heimer(presenter), B. Lowe, R. Nuggehalli, S. Sen, Creating an Testbed of Industry Problems for OR Model and Algorithm Development, INFORMS Annual Conference, San Francisco, California, USA (11/13/2005) pdf Many OR researchers are challenged with developing algorithms and models for industrial problems...without industry data. Data access can influence the applicability and impact of research. This panel session will analyze the challenges of sharing industry data and explore viable paths to creating a test bed of industry problems for OR. S. Mehrotra (presenter) , Z. Li, H. Sheng, On Generalized Branching Methods for Mixed Integer Programming, IMA Special Workshop: Mixed-Integer Programming , Minneapolis, Minnesota, USA (07/29/2005) ppt pdf We present a restructuring of the computations in Lenstra's methods for solving mixed integer linear programs. A concept of Adjoint lattices is introduced and used for this purpose. This allows us to develop branching on hyperplane algorithms in the space of original variables, without requiring any dimension reduction for pure or mixed integer programs. Reduced lattice basis are also computed in the space of original variables. Based on these results we give a new natural heuristic way of generating branching hyperplanes, and discuss its relationship with recent reformulation techniques of Aardal and Lenstra. We show that the reduced basis available at the root node has useful information on the branching hyperplanes for the generalized branch-and-bound tree. Our development allows us to give algorithms for mixed convex integer programs with properties similar to those for the mixed integer linear programs. Implementation of this algorithm in a prototype software system "IMPACT" is described. Computational results are presented on difficult knapsack and market split problems using this approach. R. Fourer, J. Ma (presenter), K. Martin, Optimization Services (OS) -- A Unified and Standard Framework for Optimization Over the Internet, IFORS Triennial Conference, Honolulu, Hawaii, USA (07/14/2005) ppt pdf We present a standard architecture under our concept of "Optimization Services" as unified framework for next generation optimization systems. Through standardization of optimization representation, communication, discovery and registration, the framework provides an open computational infrastructure for all modeling system components including modeling languages, optimization servers, clients, interfaces, analyzers, solvers and simulations. R. Fourer, J. Ma, K. Martin (presenter), OSiL: An Instance Language and API for Optimization, IFORS Triennial Conference, Honolulu, Hawaii, USA (07/14/2005) ppt pdf We present OSiL (Optimization Services Instance Language), an XML Schema for representing optimization instances including general nonlinear programming, quadratic programming, user-defined functions, and optimization via simulation. In addition, we provide open-source Java libraries that simplify the exchange of problem-instance and solution information between modeling systems and solvers. R. Fourer (presenter), J. Ma, K. Martin, Extensions to an Optimization Services Instance Language, IFORS Triennial Conference, Honolulu, Hawaii, USA (07/14/2005) pdf Optimization problems of interest today go beyond the traditional linear, integer, quadratic, and smooth nonlinear types. A language for problem instances must be extended accordingly. This presentation describes prospective extensions to OSiL, our proposed language standard, in such areas as combinatorial optimization and constraint programming, stochastic programming, and semidefinite and cone programming. R. Fourer, J. Ma, K. Martin (presenter), Optimization Services Instance Language (OSiL), Solvers, and Modeling Languages, IFORS Triennial Conference, Honolulu, Hawaii, USA (07/14/2005) ppt pdf We present OSiL (Optimization Services Instance Language), an XML Schema for representing optimization instances including general nonlinear programming, quadratic programming, user-defined functions, and optimization via simulation. In addition, we provide open-source Java libraries that simplify the exchange of problem-instance and solution information between modeling systems and solvers. R. Fourer, J. Ma (presenter), K. Martin, Optimization via Simulation under Optimization Services (OS), IFORS Triennial Conference, Honolulu, Hawaii, USA (07/14/2005) ppt pdf Optimization Services (OS) is a standard and universal framework for optimization over distributed systems. We show how OS provides the infrastructure for performing optimization via simulation, where optimization and simulation can be located anywhere over the Internet, and how the process can be parallelized. Performance and other issues are addressed. J. Ma (presenter), Optimization Services(OS), Lindo Inc., Chicago, Illinois, USA (05/18/2005) ppt pdf From Lindo's business perspective, we present a general optimization system design introduced under our new concept of Optimization Services (OS) along with its Optimization Services Protocol (OSP). Optimization Services is intended to be a unified framework for the next generation distributed optimization systems, mainly optimization over the Internet. Thus Optimization Services can be regarded as the Operations Research Internet. The corresponding Optimization Services Protocol is intended to be a set of industrial standards. J. Ma (presenter), Optimization Services(OS), T.J. Watson Lab, IBM, Yorktown Heights, New York, USA (06/23/2005) ppt pdf We present a general optimization system design introduced under our new concept of Optimization Services (OS) along with its Optimization Services Protocol (OSP). Optimization Services is intended to be a unified framework for the next generation distributed optimization systems, mainly optimization over the Internet. Thus Optimization Services can be regarded as the Operations Research Internet. The corresponding Optimization Services Protocol is intended to be a set of industrial standards. J. Ma (presenter), Optimization Services(OS), Thesis Presentation, Industrial Engineering Management Sciences, Northwestern University, Evanston, Illinois, USA (05/06/2005) ppt pdf This doctoral thesis presents a general optimization system design introduced under our new concept of Optimization Services (OS) along with its Optimization Services Protocol (OSP). Optimization Services is intended to be a unified framework for the next generation distributed optimization systems, mainly optimization over the Internet. Thus Optimization Services can be regarded as the Operations Research Internet. The corresponding Optimization Services Protocol is intended to be a set of industrial standards. J. Ma (presenter), Fundamentals of Modeling Systems and a System Approach to Simulation Optimization, Industrial Engineering and Management Sciences, Northwestern University, Evanston, Illinois, USA (02/02/2005) ppt pdf In this talk we explain the fundamentals of modeling systems and present an overview of how optimization via simulation can be integrated into the Optimization Services framework. R. Fourer, J. Ma (presenter), K. Martin, Optimization Services (OS) -- The Internet for OR and The Next Generation NEOS, Motorola Inc., Schaumburg, Illinois, USA (01/18/2005) ppt pdf From the Motorola business perspective, We present a general optimization system design introduced under our new concept of Optimization Services (OS) along with its Optimization Services Protocol (OSP). Optimization Services is intended to be a unified framework for the next generation distributed optimization systems, mainly optimization over the Internet. Thus Optimization Services can be regarded as the Operations Research Internet. The corresponding Optimization Services Protocol is intended to be a set of industrial standards. R. Fourer, J. Ma (presenter), K. Martin, A Unified XML-Based Framework for Optimization Services, 9th INFORMS Computing Society Conference, Annapolis, Maryland, USA (01/07/2005) ppt pdf I will present a general design for XML-based, service-oriented distributed architecture introduced under our concept of Optimization Services (OS) and its corresponding Optimization Services Protocol (OSP), which includes 20+ specifications of Optimization Services x Languages (OSxL), intended as unified framework for next generation (mathematical/algebraic) modeling systems over distributed systems, including the Internet. This project originated as an initiative to start a wider level of cooperation to move toward a final standardization and facilitate a healthier development environment for research in the area of Operations Research (OR) and Management Sciences (MS). Through standardization of modeling representation, communication, discovery and registration, the framework provides an open infrastructure for all modeling system components including modeling language environment, servers, registries, communications clients, interfaces, analyzers, solvers and simulations. The goal is that all the algorithmic codes will be implemented as services under this framework and customers use these computational services similar to daily utility services. Although the modeling system framework is intended as an infrastructure for the area of OR/MS, the design concept and philosophy is general enough to be learned and adopted by designers of any distributed system and architectures. G. Gassmann (presenter), Data requirements for stochastic Data requirements for stochastic solvers, 9th INFORMS Computing Society Conference, Annapolis, Maryland (01/06/2005) ppt pdf This talk gives an overview of the latest developments in an stochastic programming problem representations. R. Fourer, J. Ma (presenter), K. Martin, Optimization Services (OS) -- The Internet for OR, Industrial Engineering and Management Sciences, Northwestern University, Evanston, Illinois, USA (11/24/2004) ppt pdf What is Optimization Services (OS). It relates, affects, and benefits everyone of you and you will be the first ones to see this totally new and exiting technology for OR! In the first presentation of a series of coming Optimization Services talks, we will give you a general picture of Optimization Services made really easy to understand. You will see what the next generation NEOS is, why Optimization Services is the Internet for OR, how Optimization Services is potentially the real COIN-OR (COmputational INfrastructure for OR) and A Demo! R. Fourer, J. Ma (presenter), K. Martin, Optimization Services Instance Language (OSiL) Part I, II, Industrial Engineering and Management Sciences, Northwestern University, Evanston, Illinois, USA (11/22/2004) ppt pdf Distributed computing technologies such as Web Services are growing rapidly in importance in today’s computing environment. In the area of mathematical optimization, it is becoming increasingly common to separate modeling languages, client GUI environments and systems from optimization solvers and other algorithmic codes. In fact, the modeling language software, solver software, and data used to generate the model instance might reside on different machines using different operating systems. Such a distributed environment makes it critical to have an open standard for exchanging model instances. In this talk I present OSiL (Optimization Services Instance Language), an XML Schema for representing general mathematical optimization instances including general nonlinear, programming, network programming, quadratic programming, user-defined functions, black box non-closed form simulations, and constraint logic programming. For those of you who are familiar with MPS format for linear programming, think of OSiL as a more general and powerful representation. Since the Optimization Services instance Language is part of a much broader general framework for optimization modeling systems using Optimization Services (OS), the talk will start with a brief 10-15 minute introduction of the Optimization Services Framework to provide a better picture of the context for the OSiL Language. The Optimization Services Framework is presented in an earlier talk at the department and will show up in most of the talks in the series. Due to the one-hour limit, I will concentrate on the linear part of the OSiL and how it can easily be extended to more general optimization types. The linear part of OsiL originates from a pure linear programming format called LPFML by Fourer, Lopez, and Martin. In a possible later talk, I will give detailed extension to quadratic programming, optimization over black box simulation, constraint logic programming etc. and the library and open source API that help to read and write the OSiL instance depending on interests. R. Fourer, J. Ma (presenter), K. Martin, Optimization Services (OS) Framework and OSP Protocols (OSxL) “Combining Operations Research with Computing Technology”, INFORMS Annual Conference, Denver, Colorado, USA (10/24/2004) ppt pdf We introduce a general design for "Optimization Services", an XML-based architecture intended as a unified framework for next-generation optimization systems. Through standardization of optimization representation, communication, discovery, and registration, this framework provides an open infrastructure for all modeling system components, including modeling languages, optimization servers, clients, interfaces, analyzers, solvers and function simulators. R. Fourer, L. Lopez, K. Martin (presenter), LPFML: A W3C XML Schema for Linear and Integer Programming, INFORMS Annual Conference, Denver, Colorado, USA (10/24/2004) ppt pdf One way to encourage modeling language and solver compatibility is to use a standard representation of a problem instance. We present LPFML Schema, a W3C Schema for representing linear programming problem instances in XML. We describe a library of open-source C++ classes that facilitate the exchange of information between modeling languages and solvers. We show how these classes are used to provide previously unavailable modeling language-solver connections. LPFML is officially subsumed into the new version, OSiL (Optimization Services instance Language) by R. Fourer, J. Ma, K. Martin. R. Fourer, J. Ma (presenter), K. Martin, Optimization Services Framework and OSxL Protocols, NEOS team, Argonne National Lab, Argonne, Illinois, USA (09/14/2004) ppt pdf From the NEOS network-enabled optimization server perspecitve, We present a general optimization system design introduced under our new concept of Optimization Services (OS) along with its Optimization Services Protocol (OSP) to illustrate a blueprint for the next generation NEOS. R. Fourer (presenter), L. Lopez, K. Martin, An XML-Based Standard for Representing Linear Programming Problem Instances, Brunel University, London, UK (06/22/2004) pdf We present the LPFML Schema, a W3C Schema for representing linear programming problem instances in XML. LPFML is officially subsumed into the new version, OSiL (Optimization Services instance Language) by R. Fourer, J. Ma, K. Martin. R. Fourer, J. Ma (presenter), A General and Unified Design and Framework for Distributed Optimization, Industrial Engineering and Management Sciences, Northwestern University, Evanston, Illinois, USA (12/03/2003) ppt pdf This is the proposal for the Optimization Services thesis. It is the first conception of the Optimization Services. It lays out the original ideas and guidelines for designing and developing the Optimization Services framework.
|
|
|
|
|
|
|
|
|
|
|
|
|
| © Optimization Services (OS) 2006 | All rights reserved. | |