Contact: Daniel Fylstra
INCLINE VILLAGE, NV -- December 15, 2007 -- Premium Solver Platform Stochastic Edition, a new software product from Frontline Systems, developer of the Excel Solver and leading-edge optimization and simulation software, offers capabilities not available in any other product, at any price, for optimization of models with uncertainty. A powerful superset of Frontline's market-leading Premium Solver Platform for Excel, the Stochastic Edition is the first commercial product to support new methods of robust optimization, and the first and only product enabling users to find optimal solution to the complete spectrum of models with uncertainty, using the full spectrum of solution methods: simulation optimization, stochastic programming, and robust optimization.
Complete, Unified Framework for Models with Uncertainty
The Stochastic Edition provides one unified and easy-to-use way to define Solver models that include uncertainty. It can automatically analyze users' models to determine which solution method can be applied. Models may include:
- Uncertainties (random variables) that may appear in the objective and constraints
- Normal or first-stage ('here and now') decision variables
- Later-stage ('wait and see') decision variables, also called recourse variables
- Normal ('hard') constraints, chance ('soft') constraints, and an objective function, that may depend on the normal and recourse decision variables, and on the uncertainties
All these capabilities are needed to handle the full spectrum of optimization models with uncertainty, yet to Frontline's knowledge, no other software product inside and outside Excel provides them all. Few other optimizers handle uncertainty at all, and those that do generally provide for either chance constraints or recourse decisions, but not both.
Fast, Scalable Solutions for Common Cases; Good Solutions for All Cases
The Stochastic Edition can find good solutions - given enough time - to uncertain models of 'arbitrary' form, with non-smooth functions and decision-dependent uncertainties, using simulation optimization - even though such models are non-convex and often non-smooth global optimization problems.
But the Stochastic Edition can automatically recognize common special cases - for example stochastic linear programming models - and find proven optimal solutions at faster speeds for these models. Unlike simulation optimization models, stochastic LPs can be scaled up to model operations in large enterprises.
The Stochastic Edition can assist users in building models that involve uncertainty. It can automatically diagnose a model and point out formulas that violate the requirements of stochastic LPs, or that require the more computationally expensive approach of simulation optimization.
Pricing and Availability
Premium Solver Platform Stochastic Edition V8.0 is available now. A fully functional, free trial version can be downloaded at Frontline's Website www.solver.com. A single user license with required first-year annual support is priced at $2,995 in the U.S. A special package of Premium Solver Platform Stochastic Edition V8.0 and Risk Solver V8.0, plus annual support for both products, is priced at $4,000. Special academic prices are available for research and teaching.
About Frontline Systems, Inc.
Frontline Systems, Inc. (www.solver.com) is a leading developer of optimization and simulation software, and the leader in spreadsheet optimization software that helps analysts and managers optimally allocate scarce resources - money, equipment, and people - to realize substantial cost savings. Frontline developed the solvers/optimizers in Microsoft Excel, Lotus 1-2-3 and Quattro Pro, distributed to more than 500 million spreadsheet users. Founded in 1987, Frontline is headquartered in Incline Village, Nevada (775-831-0300 or email@example.com). Premium Solver is a trademark of Frontline Systems, Inc.
Technical Features of Premium Solver Platform Stochastic Edition
Model transformation and solution methods implemented in the Premium Solver Platform Stochastic Edition include:
- Robust optimization (RO) methods for linear programming problems with uncertainties affecting the objective and constraints. Chance constraints specify a probability of satisfaction, which is converted to a 'budget of uncertainty.' Monte Carlo simulation is used to obtain bound and shape information for the uncertainties. This information is used to automatically create a robust counterpart problem - either an LP or an SOCP - which is then solved. This method is scalable to large size (tens to hundreds of thousands of variables and constraints).
- Robust optimization methods and stochastic programming (SP) methods for two-stage stochastic linear programming problems with recourse ('wait and see') decisions. Scenarios are automatically created via built-in Monte Carlo simulation, or they may be drawn from user-defined cell ranges of sample values on the spreadsheet. With the benefit of second stage or recourse decisions, solutions are typically 'well-hedged' but not overly conservative. The RO and SP methods are scalable to large size, though scalability of the SP methods may depend on specialized Solver Engines.
- High-speed simulation optimization methods for problems more general than linear mixed-integer (non-linear, non-smooth or non-convex), where uncertainties may depend on the first-stage decisions. Normal and chance constraints may be used, and a wide range of statistical aggregates can be used to summarize uncertainty in the objective and constraints. A simulation is performed on each major iteration of the optimization. This method is very general, but computationally very expensive, and usually not scalable to large size problems.