ICSOFT-EA: Scientific Workflow

  • Posted on: 26 July 2016

This process represents a generic evolutionary algorithm simplified to facilitate its understanding. Data objects are in fact hidden and each phase is isolated in a separate Sub-Process.

We decomposed the algorithm in its main phases: Initialization, Point Generation, Evaluation, Selection and Stopping Criteria.

The inner structure of all Sub-Processes can be further exploded and modeled in more details, if necessary. The aim of this generalization is to enable the creation of as many different optimization algorithms as possible based on this simple skeleton. It is possible to model the single phases of any optimization algorithm and freely combine them thanks to the BPMN expressiveness.

This opens the door to extensive yet simple algorithm customization and hybridization.