Optimisation

From ASCEND

Optimisation problems are problems where there are more variables than constraining equations. This means that there can be large feasible regions in the variable space that solve the constraining equations. The job of an optimiser, then, is to find a a best possible feasible solution, where 'best' is defined by an objective function, which is either maximised or minimised according to what the user decides.

Available optimisers in ASCEND include

• CONOPT (closed source proprietary solver)
• IPOPT (under development, see also NonproprietaryOptimization)
• MINOS (closed source, currently broken)
• MakeMPS (an exporting interface for linear problems, currently broken)

As an aside, when doing Conditional modelling with our CMSlv, the CONOPT optimiser is currently required, and that is one of the reasons that we want to provide support for a non-proprietary optimising solver.

Syntax

For optimisation problems in ASCEND, you need to have some degrees of freedom in your model, and you need an objective statement, which can be of the form:

MAXIMIZE myvar;

or

MINIMIZE othervar;

An example of a simple complete optimisation model in ASCEND is given in Template:Models/conopttest.a4c in the ModelLibrary.