What's New in LINGO 15.0

Release 15.0 of LINGO includes support for new platforms as well as a wide range of performance enhancements and new features.

1. Native Macintosh Support:

LINGO's user interface has been entirely rewritten to offer native support for the Macintosh. Below is an image of the Mac version running a small nonlinear program on a MacBook Pro:

2. GUI Interface for Linux Versions of Lingo:

Prior to LINGO 15, Linux versions had a command-line interface, as opposed to the easier to use GUI interface found on Windows versions. Linux versions of LINGO now have a full GUI interface similar to the Windows version's interface.

3. Conic Solver Enhancements:

The Conic Solver option has been enhanced through the addition of Semi-Definite Program (SDP)/Positive Definite (POSD) capabilities using the @POSD function. As an example, if one is estimating a covariance matrix for a portfolio, it's possible to use @POSD to force the matrix to be positive semi-definite, which is a property required of any covariance matrix.

4. Linear Solver Improvements:

Simplex LP algorithm implementation has been improved for speed and robustness.
The performance improvements compared to previous version is 90% for primal simplex and 45% for the dual simplex.

5. Integer Solver Improvements:

Knapsack related cuts improvements. Significantly faster solve times on models with certain knapsack-like constraints.
Improved default node selection rules improves performance on most integer models.
New branching variable rule options: maximum coefficients and neighborhood branching, which can reduce number of branches on certain integer models.
Perspective reformulation capability gives improved performance on quadratic portfolio models with semi-continuous variables, e.g. min-buy quantities.

6. Nonlinear Solver Improvements:

Improved default settings for nonlinear models gives a 5% average speed improvement.
Faster processing of long nonlinear expressions in nonlinear models.

7. Presolver Improvements:

New preprocessing for the linear and integer solvers to significantly reduce coefficient density of certain dense matrices.

8. Other Improvements:

Quadratic repair feature for near-convex Quadratic Programs (QP). Gives improved performance on not quite convex QP’s.
Additional multithreaded code added to model generator to improve generation times on large models.
 

© Copyright 2015 Lindo Systems Inc.