LINDO Systems is proud to introduce LINGO 11.0. The new features in LINGO 11.0 include the following:

The simplex solvers include a number of new features:

• Large linear models solve an average of 80% faster using the newly enhanced dual and primal simplex solvers.
• The new release includes several enhancements to the sparse LU decomposition routines and the addition of advanced basis repairing techniques.
• Relative to earlier releases, the Simplex solvers offer improved handling of numerically difficult problems.
• Improvements in the dual simplex method that is used to reoptimize at each branch in the branch-and-bound tree on integer models, allowing more branches to be processed per second in the search tree.
• Better handling of numerically difficult linear and integer models.

LINGO's MIP solver improvements include:

• The MIP solver is now an average of 40% faster on a wide range of integer models.
• An advanced implementation of Feasibility Pump heuristic speeds the finding of feasible solutions on many difficult problems.
• An advanced implementation of the Relaxation Induced Neighborhood Search (RINS) technique finds improved integer solutions faster.
• General and local branching strategies have been improved to exploit model structure.
• Rounding techniques have been expanded to exploit a wider range of constraint structures.
• Performance has been improved on probing through out the branch-and-bound tree.
• Cut management has been improved.

The global solver includes a number of enhancements for solving non-convex, nonlinear models to global optimality:

• New discrete space constraint propagation and convex-cut generation improve performance on nonlinear models with integer variables and/or non-smooth functions.
• A multi-formulation capability improves bounding performance and enhances tractability through efficient cut generation.

The barrier solver improvements include:

• The barrier method solves large-scale linear and quadratic models an average of 20% faster.
• Improved handling of numerically difficult models.

The general nonlinear solver improvements include:

• On average, the nonlinear solver is now 25% faster on highly nonlinear and complex models.
• An efficient implementation of Algebraic Second Order Derivatives has resulted in improved speed and solution precision.

When solving linear programs on a multi-core machine, users can specify that two, or more, of LINGO's four different LP solvers are to be run in parallel on different cores. LINGO then returns the solution from the first solver to finish optimization.

LINGO's new K-Best MIP solver will return the K best solutions (based on objective value) to integer models, where K is a user controlled parameter. You may then scroll through the K solutions, examine them, and select the one of most interest based on the value of a tradeoff variable.

You may now generate the dual formulation of a linear programming model with the LINGO|Generate|Dual command.

Many models require certain variables to either be 0, or lie within some non-negative range, e.g., 10 to 20. Modeling this condition in LINGO in the past meant having to add an additional 0/1 variable and two additional constraints. LINGO now allows you to set semi-continuous variables directly with the @SEMIC statement.

You may now specify sets of binary variables that are of type SOS1 (at most.one nonzero), SOS2 (at most two nonzero and adjacent) and SOS3 (variables must sum to 1). The cardinality feature allows you to specify sets of variables where, at most, k variables can be nonzero. These functions allow the MIP solver to handle models more efficiently and cuts down on the number of constraints required in your models.

In prior releases, LINGO was only able to access external data sources using static links in the Data sections of models. The @OLE, @ODBC, @TEXT and @POINTER functions are now available in Calc sections, allowing you to dynamically build links for importing and exporting data. These import/export functions may also be placed in a loop, which is useful for solving multiple instances of a model in a loop.

The @RANK sorting function may be used in calc sections to efficiently sorts vectors and return the ranks of their members in a separate vector.

Support of the following trigonometric functions has been added: ACOS, ASIN, ATAN, ATAN2, COSH, SINH, TANH, ASINH, ACOSH, and ATANH. These functions are fully supported by the global solver, allowing for global solutions to highly non-convex, trigonometric models.

LINGO's online help files have been ported to compiled HTML for native support under Windows Vista.

In addition to numerical data, @POINTER can now import and export set members via a memory location passed by a calling application.

An example illustrating how to call LINGO from ASP.NET is now included in the \LINGO11\Programming Samples folder.

The @SOLU function in calc sections generates standard LINGO-style solution reports, @OBJBND returns the bound on the current objective, and @TIME returns the elapsed solver time.