Class AdaBoostTrainer

java.lang.Object
org.tribuo.classification.ensemble.AdaBoostTrainer
All Implemented Interfaces:
com.oracle.labs.mlrg.olcut.config.Configurable, com.oracle.labs.mlrg.olcut.provenance.Provenancable<TrainerProvenance>, Trainer<Label>
public class AdaBoostTrainer extends Object implements Trainer<Label>
Implements Adaboost.SAMME one of the more popular algorithms for multiclass boosting. Based on this paper.

If the trainer implements WeightedExamples then it performs boosting by weighting, otherwise it uses a weighted bootstrap sample.

See: 

 J. Zhu, S. Rosset, H. Zou, T. Hastie.
 "Multi-class Adaboost"
 2006.

  • Field Details

    • innerTrainer

      @Config(mandatory=true, description="The trainer to use to build each weak learner.") protected Trainer<Label> innerTrainer

    • numMembers

      @Config(mandatory=true, description="The number of ensemble members to train.") protected int numMembers

    • seed

      @Config(mandatory=true, description="The seed for the RNG.") protected long seed

    • rng

      protected SplittableRandom rng

    • trainInvocationCounter

      protected int trainInvocationCounter

  • Constructor Details

    • AdaBoostTrainer

      public AdaBoostTrainer(Trainer<Label> trainer, int numMembers)
      Constructs an adaboost trainer using the supplied weak learner trainer and the specified number of boosting rounds. Uses the default seed.
      Parameters:
      trainer - The weak learner trainer.
      numMembers - The maximum number of boosting rounds.

    • AdaBoostTrainer

      public AdaBoostTrainer(Trainer<Label> trainer, int numMembers, long seed)
      Constructs an adaboost trainer using the supplied weak learner trainer, the specified number of boosting rounds and the supplied seed.
      Parameters:
      trainer - The weak learner trainer.
      numMembers - The maximum number of boosting rounds.
      seed - The RNG seed.

  • Method Details

    • postConfig

      public void postConfig()
      Used by the OLCUT configuration system, and should not be called by external code.
      Specified by:
      postConfig in interface com.oracle.labs.mlrg.olcut.config.Configurable

    • toString

      public String toString()
      Overrides:
      toString in class Object

    • train

      public Model<Label> train(Dataset<Label> examples, Map<String, com.oracle.labs.mlrg.olcut.provenance.Provenance> runProvenance)
      If the trainer implements WeightedExamples then do boosting by weighting, otherwise do boosting by sampling.
      Specified by:
      train in interface Trainer<Label>
      Parameters:
      examples - the data set containing the examples.
      runProvenance - Training run specific provenance (e.g., fold number).
      Returns:
      A WeightedEnsembleModel.

    • getInvocationCount

      public int getInvocationCount()
      Description copied from interface: Trainer
      The number of times this trainer instance has had it's train method invoked.

      This is used to determine how many times the trainer's RNG has been accessed to ensure replicability in the random number stream.

      Specified by:
      getInvocationCount in interface Trainer<Label>
      Returns:
      The number of train invocations.

    • getProvenance

      public TrainerProvenance getProvenance()
      Specified by:
      getProvenance in interface com.oracle.labs.mlrg.olcut.provenance.Provenancable<TrainerProvenance>