Documentation
/ User's manual in Python
: 
| Chapter VIII. The Relauncher module | ||
|---|---|---|
 |  | |
Table of Contents
The aim of the Relauncher module is to provide a general architecture for all parametric study and is, because of this
generality aspect, heavily used throughout the Uranie platform. However, it is generally used for more advanced
techniques than the usual recommended first steps and it allows more flexible distribution approaches.
Studies
allowed thanks to this module (no concrete study will indeed be described in this chapter, as the module is more a
support for many other classes in other modules) aim at evaluate a model for different input parameter's values and
check the evolution of its outputs. These studies can be split into two kinds:
the opened-loop ones: all input parameters are known at the start (Monte-Carlo simulation for instance).
the closed-loop ones: results of the evaluation will impact the next input parameter's value (optimisation for instance).
One can find examples of how to run analysis with the relauncher implementation in Section XIV.9.
Item evaluations can be time consuming, and many kinds of such studies are able to distribute them on computer resources. This architecture provides different ways to use these resources. However, if evaluation is fast, evaluation distribution is counterproductive.
Because of the very specific organisation of this module, the class hierarchy is not shown here but split into pieces which will be introduced in the next section. Every component is discussed in more details in the following sections and a schematic description of the needed steps to define a relauncher procedure is shown in Figure VIII.1.
| | | |
| VII.3. Multicriteria optimisation |  | VIII.2. Relauncher abstraction levels |