What's New in LINGO 10.0

The new release of LINGO includes a wide range of solver enhancements that improve performance and robustness and a number of new features that make expressing and solving models easier and more flexible.

Improved MIP Solver

Improvements in the integer solver include the following:

• Faster MIP presolving
• Enhancements to exploit Special Ordered
• Sets and Semi-Continuous variables in cut generation, presolving and branching
• Improved handling of BigM coefficients
• Improvements in cut generation strategies to reduce B&B size

Improved Global Solver

Improvements in the global solver include the following:

• Auto model re-expression for faster local
• search and tighter convex relaxations
• Mathematical structure recognition to derive dedicated cuts and resort to more appropriate/faster solvers
• Improved preprocessing to derive implied variable bounds and types

Faster Dual Sover

The dual solver now performs 40% faster than the dual solver in previous releases.

Improved Nonlinear Solver

Improvements in the nonlinear solver include the following:

• A new option to compute second-order derivatives analytically
• Faster first order derivative calculations

Scripting Capability

The syntax of the calc section has been extended to include full scripting capabilities. This allows you to solve one or more models in a loop, with model solutions being fed to subsequent models as input. Some of the control structures and functions added to support scripting include:

• The SUBMODEL statement allows you to define one or more submodels within a main model. Submodels may be solved using the @SOLVE command.
• @IFC/@ELSE statements for conditional IF/THEN/ELSE branching.
• @WHILE statement for executing compound statements in a loop.
• Refer to sample models LOOPPORT, LOOPCUT and LOOPTSP for examples of using scripting loops to solve, respectively, cutting stock, Markowitz portfolio and traveling salesman models.

Debugging Nonlinear and Integer Models:

Starting with release 9.0, LINGO had the ability to debug infeasible and unbounded models linear models. LINGO's debugger can now debug the full range of model classes, including mixed integer linear, quadratic, nonlinear and mixed integer nonlinear models

Improved Error Tracking

LINGO has always had the ability to pinpoint compiler errors in the model source. Unfortunately, runtime errors in models were not pinpointed with the same degree of accuracy. LINGO can now trace runtime errors back to a single source line of the model, greatly facilitating model development and debugging.

IMPROVED FUNCTIONS

@TABLE Output Function

The @TABLE function displays either an attribute's values or a set's members in tabular format. @TABLE can also display sets and attributes of more than two dimensions, and provides you with great control over how multidimensional objects get displayed. You can refer to either the QUEENS8.LG4 or PERT.LG4 models for examples of @TABLE.

@NORMSINV() Function

LINGO now supports the @NORMSINV() function for computing the inverse of the standard normal cumulative distribution.

Large Address Aware

Win32 versions of LINGO are now "large address aware,"allowing access to an additional gigabyte of memory, for a total address space of 3GB.

JNI Interface

A Java JNI interface is now available for the LINGO callable DLL. A Java staff scheduling example may be found in \Lingo10\Programming Samples\Java.

64-bit Version:

64-bit versions of LINGO are now available for the AMD Athlon and Intel Xeon processors running Windows XP64. These versions take full advantage of the 64-bit address space, allowing access to virtually unlimited memory.

Additional Options

The following additional Options have been added to Lingo 10:

• Priority of unary minus
• Linear generation mode for reduced memory consumption
• Linear optimality tolerance
• Second order derivatives
• Default lower variable bound
• BigM coefficient threshold

SQL Server Example

An example in now included on using the @ODBC() function to interface with Microsoft SQL Server.

© Copyright 2006 Lindo Systems Inc.