Type Members

1.  trait Learner extends Serializable

   Created by maxhutch on 11/14/16.

2.  trait Model[+T <: PredictionResult[Any]] extends Serializable

   Created by maxhutch on 11/14/16.

3.  trait MultiTaskLearner extends Serializable

   A learner that trains on multiple labels, outputting a single model that makes predictions for all labels.

4.  trait MultiTaskModel extends Model[MultiTaskModelPredictionResult]

   A model that predicts a sequence of values, corresponding to multiple labels.

5.  trait MultiTaskModelPredictionResult extends PredictionResult[Seq[Any]]

   Container for predictions made on multiple labels simultaneously.

6.  trait MultiTaskTrainingResult extends TrainingResult
7.  class ParallelModels extends MultiTaskModel

   A set of parallel models, one for each label.

8.  class ParallelModelsPredictionResult extends MultiTaskModelPredictionResult

   A container that holds the predictions of several parallel models for multiple labels.

9.  trait PredictionResult[+T] extends AnyRef

   Container for prediction results; must include expected values Created by maxhutch on 11/29/16.

10.  trait RegressionResult extends PredictionResult[Double]

    Additional regression-specific interface

    Additional regression-specific interface

    This interface is experimental and SHOULD BE REVIEWED before being merged into master. In particular, an explanation of how the different methods relate to each other, how predictive uncertainty is decomposed, and what the assumptions are should be added, as these are currently not entirely clear.

    For example, does the interface assume that the predictions are the mean of a predictive distribution (as opposed to, for example, the median, or the value with highest probability)? Does it assume the predictive distribution to be normal? Such assumptions are fine, but should be explicitly stated.

11.  trait TrainingResult extends Serializable

    Created by maxhutch on 12/4/16.