Developers: adding a new preference learning algorithm - TAPeri/pl-toolbox GitHub Wiki

When adding a new algorithm to the tool, you should create a dedicated package under plt.plalgorithm. Within this package, you should include at least three classes:

• Extend plt.algorithm.Model: this abstract class provides the basic routines to use a preference learning model. When creating a new model, you will have to define a method to calculate the outputs of the model (calculatePreference), a method to save the model in a file (save) and a method to copy the model (clone). is used to query the outputs of a preference model and save it to a file
• Extend plt.algorithm.PLAlgorithm: this abstract class provides basic access to set up and run a training algorithm. When creating a new algorithm, you will have to define a method to test the correctness of the parameters selected by the user (testParameters), a method to initialise the training procedure (beforeRun), a method to run the training process (run) and a method to obtain a textual description of the algorithm parameters (getProperties).
• Implement plt.gui.algorithms.GUIconfigurator: this interface defines the connection between the GUI and the preference learning algorithm. When creating a new algorithm, you will have to define a method that creates the necessary GUI elements to set up all parameters of the algorithm (ui) and a method that initialises the algorithm using the parameters given through the GUI (algorithm).

In addition to those three classes, you will have to modify the class plt.gui.algorithms.SupportedAlgorithms: this class contains the label given to the algorithms on the GUI and connects it to the relevant extension of the class plt.gui.algorithms.GUIconfigurator.