Package org.tribuo.classification.sequence.viterbi

Class ViterbiTrainer

java.lang.Object
org.tribuo.classification.sequence.viterbi.ViterbiTrainer
All Implemented Interfaces:
com.oracle.labs.mlrg.olcut.config.Configurable, com.oracle.labs.mlrg.olcut.provenance.Provenancable<TrainerProvenance>, SequenceTrainer<Label>
public final class ViterbiTrainer extends Object implements SequenceTrainer<Label>
Builds a Viterbi model using the supplied Trainer. Has a parameter to control the label features which are added to the features supplied by the data.

  • Constructor Details

    • ViterbiTrainer

      public ViterbiTrainer(Trainer<Label> trainer, LabelFeatureExtractor labelFeatureExtractor, ViterbiModel.ScoreAggregation scoreAggregation)
      Constructs a ViterbiTrainer wrapping the specified trainer, with an unbounded stack size.
      Parameters:
      trainer - The trainer to wrap.
      labelFeatureExtractor - The feature extraction function for labels.
      scoreAggregation - The score aggregation function.

    • ViterbiTrainer

      public ViterbiTrainer(Trainer<Label> trainer, LabelFeatureExtractor labelFeatureExtractor, int stackSize, ViterbiModel.ScoreAggregation scoreAggregation)
      Constructs a ViterbiTrainer wrapping the specified trainer.
      Parameters:
      trainer - The trainer to wrap.
      labelFeatureExtractor - The feature extraction function for labels.
      stackSize - The stack size.
      scoreAggregation - The score aggregation function.

  • Method Details

    • train

      public SequenceModel<Label> train(SequenceDataset<Label> dataset, Map<String,com.oracle.labs.mlrg.olcut.provenance.Provenance> runProvenance)
      The viterbi train method is unique because it delegates to a regular Model train method, but before it does, it adds features derived from preceding labels. The pipeline upstream of this call should not care that these features are being added - that is, we would not want to make the upstream logic worry about what kind of trainer will be used and have conditional logic that says to add special label-derived features if using the ViterbiTrainer. So, these one-of-a-kind unique-in-the-world label-derived features are generated here and added to the sequence examples of the passed in dataset. If you pass in a MutableSequenceDataset, then please be aware that your dataset will be modified after calling this method and therefore subsequent calls to other SequenceModel.train methods with your dataset should be avoided. If you pass in an ImmutableSequenceDataset, then please be aware that your entire dataset is going to be copied as a MutableSequenceDataset - so there is a memory penalty.
      Specified by:
      train in interface SequenceTrainer<Label>
      Parameters:
      dataset - The input dataset.
      runProvenance - Any additional information to record in the provenance.
      Returns:
      A SequenceModel using Viterbi wrapped around an inner Model.

    • getInvocationCount

      public int getInvocationCount()
      Description copied from interface: SequenceTrainer
      Returns the number of times the train method has been invoked.
      Specified by:
      getInvocationCount in interface SequenceTrainer<Label>
      Returns:
      The number of times train has been invoked.

    • toString

      public String toString()
      Overrides:
      toString in class Object

    • getProvenance

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