Faster and More Flexible Solvers

 

The recently released What'sBest! 8.0 solves broad classes of problems substantially faster and includes expanded and improved support for several Excel functions.

 

Faster Linear Solvers

- The new release is, on average, five times faster than the previous release on large linear models using default settings. The Barrier solver averages over four times faster. Enhancements include a new two-phase implementation in the Dual Simplex solver and, in the Barrier solver, improved exploitation of sparsity, better handling of dense columns, dynamic scaling of quadratic models, and improved exploitation of individual hardware platforms.

 

Improved Integer Solver

- Enhancements boost speed on broad classes of integer problems and improve efficiency. Memory management of branch-and-bound tree has been made more efficient. The Mixed Integer Programming presolver has been made faster on models with many continuous variables. Probing now occurs during branch-and-bound rather than just during presolve. Branching strategies have been extended to exploit decomposition structure for dramatic improvements on MIP models with strong block structure.

 

Improved Global Solver

- The Global solver includes a number of significant enhancements that make it, on average, more than twice as fast on difficult models, and it solves many models previous releases could not solve. The nonlinear/nonsmooth solution heuristic has been enhanced to find better solutions faster and with more precision. The cut generation in the convex relaxation and the global branch-and-bound scheme have been improved. The numerical stability has been increased around undefined points. Support has been added for a number of new nonlinear functions.

 

Enhanced Nonlinear Solver

- The NLP solver has been redesigned to improve solution speed. Extensions to the base GRG method, such as preprocessing, several new crash options, efficient exploitation of linearity, expanded sequential linear programming (SLP) and a sequential quadratic programming (SQP) components make the NLP solver more efficient on a wide range of models.

 

More Efficient Linear and Nonlinear Presolving

- The Presolver can help reduce the size of the original model and shorten the solution time. With this release the Presolver was completely redesigned, and support was expanded to include nonlinear and quadratic models.

 

Additional and Enhanced Excel Function Support

- What'sBest! has added support for several Excel functions including VLOOKUP, HLOOKUP, COSH, ACOSH, SINH, ASINH, TANH and ATANH. Support for SUMIF and LOG functions has been improved.

 

Allows Greater User Control

- This release of What'sBest! allows the user greater control over the solution process. Users have the option to change more of the solver's parameters, settings and tolerances. This can allow you to tailor the solver's algorithms to your particular model structure to get maximum speed. For integer models, the user can turn on and off new cut generation options in an attempt to take advantage of the model structure and reduce solution time. For nonlinear models, there is a new strategy available for generating a good starting point in the solution process.

Greater Speed in Extracting the Model

- The process of verifying and extracting the optimization information from Excel has been redesigned to be much more efficient. On large models this can dramatically shorten the setup time and allow the solver to begin iterating toward a solution much more quickly.

 

For general information on What'sBest!'s capabilities, visit our What'sBest! product page. To try a limited capacity version of What'sBest!, see our download page. To order a copy of What'sBest! or update an older version, go to our order page.