AMPLDev SP Edition
AMPLDev SP Edition is an integrated modelling and solving environment for stochastic programming and robust optimisation. It is designed to build and solve and analyse such models in a compact and non-factitious way. The focus on stochastic programming problems gives several advantages over the use of a general modelling system, such as the ability to use decomposition methods to solve large scale problems and user friendly ways to define the tree structure and eventual (integrated) chance constraints. The support for robust optimisation problems allows the user to formulate various robust formulations without knowledge of the mathematical details of the reformulation, thus focusing on the model at hand.
Optimisation under Uncertainty
AMPLDev SP can be used to investigate a large family of models for optimisation under uncertainty, including:
- Chance Constrained problems
- Integrated Chance Constrained problems
- Two-stage Stochastic Programming problems
- Multistage Stochastic Programming problems
- Robust Optimisation problems
- Soyster formulation
- Ben-Tal and Nemirovski formulation
- Bertsimas and Sim formulation
The modelling subsystem of AMPLDev SP Edition is based on our stochastic programming extensions SAMPL, which extend the leading algebraic modelling language AMPL. By combining natural definitions of the randomness of the problems with the existing features of these optimisation systems, such extensions introduce powerful constructs for formulating complex stochastic programming, (integrated) chance constrained programming models and robust optimisation problems. The modelling subsystem is able to generate model data in SMPS format, giving AMPLDev SP edition the ability to link any external solver which supports this standard.
Closely coupled with the modelling system, AMPLDev SP includes a stochastic solver FortSP, which incorporates alternative solution algorithms for SP problems, including:
- Benders' decomposition
- Level decomposition
- Lagrangean relaxation
The solver is also capable of computing good discrete feasible solutions to "real world" instances of mixed integer SP models. Deterministic equivalent instances may be also constructed and solved using the Interior Point Method (IPM).
AMPLDev SP edition is sufficiently versatile and allows the modeller to perform scenario analysis, analysis of 'Here and Now' and the 'expected value' solutions. Stochastic information such as the expected value of perfect information (EVPI) and the value of stochastic solution (VSS) are easily computed starting from the same core model. By supporting the ODBC standard for database connection, commercial systems can be used to link AMPLDev SP edition with scenario generators as well as to store and analyse the application data. The user can also take advantage of multidimensional data viewers, like On-Line Analytical Processing (OLAP) tools, for the analysis of the model data and the corresponding (optimal) solutions.