CategoricalHinge

public class CategoricalHinge

Computes the categorical hinge loss between labels and predictions.

loss = maximum(neg - pos + 1, 0) where neg=maximum((1-labels)*predictions) and pos=sum(labels*predictions)

labels values are expected to be 0 or 1.

Standalone usage: 

    Operand<TFloat32> labels =
        tf.constant(new float[][] { {0, 1}, {0, 0} });
    Operand<TFloat32> predictions =
        tf.constant(new float[][] { {0.6f, 0.4f}, {0.4f, 0.6f} });
    CategoricalHinge categoricalHinge = new CategoricalHinge(tf);
    Operand<TFloat32> result = categoricalHinge.call(labels, predictions);
    // produces 1.4

Calling with sample weight: 

    Operand<TFloat32> sampleWeight = tf.constant(new float[] {1f, 0.f});
    Operand<TFloat32> result = categoricalHinge.call(labels, predictions, sampleWeight);
    // produces 0.6f

Using SUM reduction type: 

    CategoricalHinge categoricalHinge = new CategoricalHinge(tf, Reduction.SUM);
    Operand<TFloat32> result = categoricalHinge.call(labels, predictions);
    // produces 2.8f

Using NONE reduction type: 

    CategoricalHinge categoricalHinge =
        new CategoricalHinge(tf, Reduction.NONE);
    Operand<TFloat32> result = categoricalHinge.call(labels, predictions);
    // produces [1.2f, 1.6f]

Inherited Fields

Public Constructors

CategoricalHinge(Ops tf)
Creates a Categorical Hinge Loss using getSimpleName() as the loss name and a Loss Reduction of REDUCTION_DEFAULT
CategoricalHinge(Ops tf, Reduction reduction)
Creates a Categorical Hinge Loss using getSimpleName() as the loss name
CategoricalHinge(Ops tf, String name, Reduction reduction)
Creates a Categorical Hinge

Public Methods

<T extends TNumber> Operand<T>
call(Operand<? extends TNumber> labels, Operand<T> predictions, Operand<T> sampleWeights)
Generates an Operand that calculates the loss.

Inherited Methods

Public Constructors

public CategoricalHinge (Ops tf)

Creates a Categorical Hinge Loss using getSimpleName() as the loss name and a Loss Reduction of REDUCTION_DEFAULT

Parameters
tf the TensorFlow Ops

public CategoricalHinge (Ops tf, Reduction reduction)

Creates a Categorical Hinge Loss using getSimpleName() as the loss name

Parameters
tf the TensorFlow Ops
reduction Type of Reduction to apply to the loss.

public CategoricalHinge (Ops tf, String name, Reduction reduction)

Creates a Categorical Hinge

Parameters
tf the TensorFlow Ops
name the name of the loss
reduction Type of Reduction to apply to the loss.

Public Methods

public Operand<T> call (Operand<? extends TNumber> labels, Operand<T> predictions, Operand<T> sampleWeights)

Generates an Operand that calculates the loss.

Parameters
labels the truth values or labels
predictions the predictions
sampleWeights Optional sampleWeights acts as a coefficient for the loss. If a scalar is provided, then the loss is simply scaled by the given value. If SampleWeights is a tensor of size [batch_size], then the total loss for each sample of the batch is rescaled by the corresponding element in the SampleWeights vector. If the shape of SampleWeights is [batch_size, d0, .. dN-1] (or can be broadcast to this shape), then each loss element of predictions is scaled by the corresponding value of SampleWeights. (Note on dN-1: all loss functions reduce by 1 dimension, usually axis=-1.)
Returns
  • the loss