Skip to content
Snippets Groups Projects
  1. Oct 02, 2018
  2. Sep 15, 2018
  3. Sep 13, 2018
  4. Aug 01, 2018
    • Alexander Diewald's avatar
      DSE-Process: Add an ExplorationFeature service. · fdd0872a
      Alexander Diewald authored
      
      * The IExplorationFeatureService allows to add extensions to the DSE
        process that consist of sub-modules, e.g., Variables, Evaluations,
        or additional operations on existing encodings.
      * While the IExplorationFeatureService is used to manage all existing
        extensions, the activated ones (configuration) is passed to the DSE
        by the exploration backend. Currently, the activated features are hard
        coded, but shall be moved to the UI.
      * Consequently, the IExplorationModuleService has bee removed. Its
        functionality has been moved to the ExplorationFeature abstract class
        that are single features registered with the
        IExplorationFeatureService.
      * The activated features now govern the installation of the sub-modules
        (Creators, Decoders), instead of the previous
        IExplorationModuleService.
      * When setting up the Opt4J exploration, the activated decoder
        configuration is now checked for consistency to avoid circular
        dependencies and duplicate responsibilities of decoders to create a
        phenotype.
      * An IExplorationFeature type is added to the MOEA ecore model to allow
        referencing exploration features from EMF classes.
      
      Signed-off-by: default avatarAlexander Diewald <diewald@fortiss.org>
      fdd0872a
  5. Jun 26, 2018
    • Alexander Diewald's avatar
      DSE-DI: Cleanup for DI-based decoders (DI --> Guice). · 540c157b
      Alexander Diewald authored
      
      * Create a GuiceDecoder class that serves as a container at which
        Guice-based decoders can be registered. It calls the Decoder that
        provides the demanded output encoding (Provides annotation). The
        Guice injector will be responsible to resolve the dependencies between
        the decoders.
      * Simplify the Phenotype classes such that there exists only a single
        phenotype that can host sub-phenotypes.
      * Also remove the generic from the phenotype that declares it's primary
        input genotype: There is almost no gain in terms of safety/robustness
        versus a ton of code complications.
      
      Signed-off-by: default avatarAlexander Diewald <diewald@fortiss.org>
      540c157b
  6. Apr 17, 2018
Loading