Solving Collections of Data Related Models within GAMS

In many applications, optimization of a collection of problems is required where each problem is structurally the same, but in which some or all of the data defining the instance is updated. Such models are easily specified within modern modeling systems, but have often been slow to solve due to the time needed to regenerate the instance, and the inability to use advance solution information (such as basis factorizations) from previous solves as the collection is processed. We describe a new language extension, GUSS, that gathers data from different sources/symbols to define the collection of models (called scenarios), updates a base model instance with this scenario data and solves the updated model instance and scatters the scenario results to symbols in the GAMS database. We demonstrate the utility of this approach in three applications, namely data envelopment analysis, cross validation and stochastic dual dynamic programming. The language extensions are available for general use in all versions of GAMS starting with release 23.7.