Linear programming, convex quadratic programming, mixed integer programming , and stochastic programming problems
The IBM Optimization Solutions and Library product family includes a set of stand-alone
solvers for LP, QP, MIP, and Stochastic programming, and a subroutine library for creating
custom applications:
For linear programming, primal and dual versions of the simplex method are implemented. The user is allowed to adjust the pricing strategy. In addition, three interior-point solvers (one primal, one primal-dual path-following, and one primal-dual predictor-corrector) are available. A regularized LP interior point option is also provided. For network linear programming, a specialized simplex method is used.
The quadratic programming solver uses two-phase approach. In the first phase, successive linear programming approximations to the actual problem are used to generate approximate solutions. When these solutions are sufficiently close, the second phase --- a standard active-set technique --- is activated. An extension to the regularized LP solver is available to provide interior point algorithms for solving quadratic programming problems.
A branch-and-bound technique is used for mixed integer programming. A simplex method is used to solve the linear programming subproblems.
The stochastic programming solver provides the capability to specify, store and retrieve, and modify stochastic scenarios and distributions, and to invoke an L-shaped Bender's decomposition solver, then to access and retrieve the solution.
OSL is available for a variety of hardware environments, from IBM compatible PCs with Windows NT or 95; to workstations from IBM, Hewlett-Packard, Silicon Graphics, and Sun; to IBM mainframes.
Input can be passed to OSL routines through array arguments in the subroutine calling sequence. OSL also has input-output service routines that read data from MPS and Lotus 1-2-3 files. Stochastic data can be read from SMPS format files.
An OSL application program may be written in C/C++ or FORTRAN.
A vendor program is available to application developers who embed OSL into products for
sale to end-users. Contact osl@vnet.ibm.com for information
Visit the OSL V2 website, the OSL Research homepage, or contact:
OSL Development IBM Corporation, 33DA MS P356 Phone: (800) 325-2733 Fax: (914) 435-5094 osl@vnet.ibm.com
Volume 31 (1992) of the IBM Systems Journal contains eight articles related to OSL. An overview of the product may be found in the article cited below.
D. G. Wilson, A brief introduction to the IBM Optimization Subroutine Library, SIAG/OPT Views and News 1 (1992), pp. 9--10.
[ Optimization Software Guide | OTC Home Page | NEOS Server | NEOS Guide ]