AIMMS - New Features AIMMS 3.8

Paragon has released the newest version of the AIMMS modeling system: AIMMS 3.8. Based on the newest insights from research & academia, requests from our customers, the latest technological developments and our personal ideas and experience, the following major features have been implemented:

Benders Decomposition for Stochastic Programming

Following the introduction in AIMMS 3.7 of language features to formulate and generate the deterministic equivalent of multi-stage stochastic recourse models, AIMMS 3.8 includes an algorithm to solve stochastic models using a stochastic Benders approach and a module to visualize stochastic scenario trees.

The nested Benders decomposition algorithm is a popular algorithm for solving (linear) multi-stage stochastic programming problems. AIMMS offers nested Benders decomposition as a user-customizable procedure inside the AIMMS library for operations on Generated Mathematical Programs (GMP library), similar to the AIMMS Outer Approximation solver.

Benefits: The Benders decomposition algorithm is faster for stochastic models than solving the deterministic equivalent with an LP solver. The open implementation allows modelers to fine-tune it to their specific stochastic model and to use advanced features of the GMP library to solve multiple sub-problems in parallel, which may drastically reduce the total solution time.

Nonlinear Presolve

AIMMS 3.8 offers a powerful presolve procedure for Nonlinear Programs and Mixed-Integer Nonlinear Programs, analogous to the presolve procedures employed by LP/MIP solvers. The AIMMS nonlinear presolve procedure applies techniques like:

Inversion of nonlinear expressions
Solving the variables in the pre-triangular part of the matrix
Tightening the bounds of variables based on the linear and nonlinear constraints

Benefits: The presolve procedure may simplify the model sent to a nonlinear solver (CONOPT, SNOPT, KNITRO, LGO, BARON, MINOS, or the AIMMS Outer Approximation solver). With the presolve, nonlinear problems may be solved faster and the solution may be better. The solver may even find a feasible solution for problems declared infeasible without the presolve, and in other cases provide proof that it really is infeasible.

Multi-Start Solve

AIMMS 3.8 includes a multi-start algorithm for nonlinear programming problems. The multi-start algorithm randomly generates starting points, groups these into clusters, calls an NLP solver for each cluster and reports back the best feasible solution as its final solution. The algorithm can also report all feasible solutions found. The AIMMS multi-start algorithm is a user-customizable procedure within the GMP library, allowing modelers to adapt the algorithm to their own needs.

Benefits: The multi-start algorithm may increase the chance of finding a good final solution. Multi-start increases the total solving time, but this may be limited by using the advanced features of the GMP library such as parallel/distributed solver sessions to solve problems with different starting points in parallel.

CPLEX 11

AIMMS 3.8 can be linked to the newest version of the widely used CPLEX solver: CPLEX 11. This LP/MIP/QP solver has added a new dynamic search algorithm, with innovative integration and sequencing of the usual branching, nodes and cuts in branch-and-cut algorithms. CPLEX 11 delivers significant performance gains, with solve times that can be up to 10 times faster for large, complex MIP business problems.

Network Object Extensions

The graphical object to display networks has been extended to include:

Curved arcs (see picture)
Segmented arcs with either 2 or 3 segments
Intuitive navigating and zooming (see picture)

GIS Extensions

The link to Geographic Information Systems (GIS) has been extended with:

More map formats, such as:
- ESRI® shape files.
- Yahoo!® Maps (see picture).
- File image based sources.
Indexed sources and layers to allow data-driven specification of multiple sources and layers in a map.

Pivot Table Extensions

The versatility of the AIMMS pivot table will be further enhanced with:

Sorting indices in the row and column areas.
Nested sorting in the grid area.
Column widths resizable by end-user.
Column widths controllable with a model parameter.

Multi-Developer Library Extensions

Working with multiple developers on a single AIMMS project using separate libraries has been made much easier with several new features, such as:

Read/write security settings for each library independently
A cross-reference tool, to analyze the references between libraries (variables, parameters, pages, etc).

Improved Web Services and Platform Portability

The AIMMS Web Services have been made more robust with new, platform-independent RPC technology. Setting up a Web Service has become much easier. Several features of AIMMS that were previously only available for the Windows version of AIMMS are now also available in the Linux x86 version of AIMMS 3.8:

Web services
Multi-agent technology
AIMMS component technology comparable to the AIMMS COM object under Windows
Remote solver sessions, allowing large stochastic models or multi-start problems to be solved in a distributed manner

AIMMS Viewer and Packed project file

At the same time as the release of AIMMS 3.8, we have also released a new product: a freely distributable AIMMS Viewer. The Viewer allows anyone to browse the results of an AIMMS-based optimization application through the graphical user interface of the application. Further details about the functionality and benefits of the Viewer can be found here.

AIMMS now also offers the possibility to pack all files related to an AIMMS project into a single, compressed file. This makes it very easy to distribute a complete AIMMS project including optimization results, for example to users of the AIMMS Viewer.

This page as PDF.

Read on: