-
Notifications
You must be signed in to change notification settings - Fork 1
Architecture

GIMME is divided in several sub-modules. The relations between them are illustrated above: the class Adaptation uses ConfigsGenAlg to select an adequate group organization in each call; ConfigsGenAlg uses RegressionAlg to predict new player states and evaluate the quality of a configuration; Finally, all submodules use the auxiliary structures InteractionsProfile, PlayerCharacteristics, PlayerState and PlayerStateGrid. Alongside, Model Bridges allow transparent model connections whenever data needs to be fetched.
Note: Throughout our schematic representations we will use some visibility notation for attributes and methods in classes. Even so, these are purely conceptual although they should be considered for good practice. Keep in mind that, due to programming limitations, attributes and methods are currently fully visible in each class.
Adaptation
Group Configuration Generation
- ConfigsGenAlg
- RandomConfigsGenAlg
- PureRandomSearchConfigsGenAlg
- EvolutionaryConfigsGenAlg
- ODPIPConfigsGenAlg (exact)
- CLinkConfigsGenAlg (legacy)
Preferences Estimation
Quality Evaluation Algorithms
- QualityEvalAlg
- Group-Based Quality Evaluation:
- Regression-Based Quality Evaluation:
- Tabular Quality Evaluation:
Auxiliary Structures
- InteractionsProfile
- PlayerCharacteristics
- PlayerState
- Personality (Inherent Preference):
- PlayerStatesDataFrame
Model Bridges
Player Data Trim