Interface SGDObjective<T>

Type Parameters:
T - The type of the output at training time.
All Superinterfaces:
com.oracle.labs.mlrg.olcut.config.Configurable, com.oracle.labs.mlrg.olcut.provenance.Provenancable<com.oracle.labs.mlrg.olcut.provenance.ConfiguredObjectProvenance>
All Known Subinterfaces:
LabelObjective, MultiLabelObjective, RegressionObjective
All Known Implementing Classes:
AbsoluteLoss, BinaryCrossEntropy, Hinge, Hinge, Huber, LogMulticlass, SquaredLoss
public interface SGDObjective<T> extends com.oracle.labs.mlrg.olcut.config.Configurable, com.oracle.labs.mlrg.olcut.provenance.Provenancable<com.oracle.labs.mlrg.olcut.provenance.ConfiguredObjectProvenance>
An interface for a loss function that can produce the loss and gradient incurred by a single prediction.

  • Method Summary

    Modifier and Type
    Method
    Description
    com.oracle.labs.mlrg.olcut.util.Pair<Double, SGDVector>
    lossAndGradient(T truth, SGDVector prediction)
    Scores a prediction, returning the loss and a vector of per output dimension gradients.

    Methods inherited from interface com.oracle.labs.mlrg.olcut.config.Configurable

    postConfig

    Methods inherited from interface com.oracle.labs.mlrg.olcut.provenance.Provenancable

    getProvenance

  • Method Details

    • lossAndGradient

      com.oracle.labs.mlrg.olcut.util.Pair<Double, SGDVector> lossAndGradient(T truth, SGDVector prediction)
      Scores a prediction, returning the loss and a vector of per output dimension gradients.
      Parameters:
      truth - The true output.
      prediction - The prediction for each dimension.
      Returns:
      The score and per dimension gradient.