PAM---A Practitioner's Approach to Modeling

Many Linear Programming (LP) practitioners formulate and document their models with block schematics of LP matrices. They subscribe to neither the column-wise emphasis of commercial matrix generators nor to the constraint orientation of recently proposed "modelling languages." PAM is a nonprocedural modeling language in which the practitioner describes an LP matrix with a set of two-dimensional tables: a table representing the block schematic of the matrix and supporting data tables. A computer program processes these tables to produce an LP matrix ready for optimization. This paper describes block-wise model formulation, discusses the relationships of the data to the model, presents an overview of PAM's matrix generation language, and gives some experience with a prototype system.