CosineSimilarity

public class CosineSimilarity

Computes the cosine similarity between labels and predictions.

Note that it is a number between -1 and 1. When it is a negative number between -1 and 0, 0 indicates orthogonality and values closer to -1indicate greater similarity. The values closer to 1 indicate greater dissimilarity. This makes it usable as a loss function in a setting where you try to maximize the proximity between predictions and targets. If either labels or predictions is a zero vector, cosine similarity will be 0 regardless of the proximity between predictions and targets.

loss = -sum(l2Norm(labels) * l2Norm(predictions))

Standalone usage:

    Operand<TFloat32> labels =
        tf.constant(new float[][] { {0.f, 1.f}, {1.f, 1.f} });
    Operand<TFloat32> predictions =
        tf.constant(new float[][] { {1.f, 0.f}, {1.f, 1.f} });
    CosineSimilarity cosineLoss = new CosineSimilarity(tf);
    Operand<TFloat32> result = cosineLoss.call(labels, predictions);
    // produces -0.5
 

Calling with sample weight:

    Operand<TFloat32> sampleWeight = tf.constant(new float[] {0.8f, 0.2f});
    Operand<TFloat32> result = cosineLoss.call(labels, predictions, sampleWeight);
    // produces -0.0999f
 

Using SUM reduction type:

    CosineSimilarity cosineLoss = new CosineSimilarity(tf, Reduction.SUM);
    Operand<TFloat32> result = cosineLoss.call(labels, predictions);
    // produces -0.999f
 

Using NONE reduction type:

    CosineSimilarity cosineLoss = new CosineSimilarity(tf, Reduction.NONE);
    Operand<TFloat32> result = cosineLoss.call(labels, predictions);
    // produces [-0.f, -0.999f]
 

Constants

int

DEFAULT_AXIS

Fields

public static final Reduction

DEFAULT_REDUCTION

Inherited Fields

From class org.tensorflow.framework.losses.Loss

public static final Reduction

REDUCTION_DEFAULT

Public Constructors

	CosineSimilarity(Ops tf) Creates a Cosine Similarity Loss using getSimpleName() as the loss name, an axis of DEFAULT_AXIS, and a Loss Reduction of DEFAULT_REDUCTION
	CosineSimilarity(Ops tf, String name) Creates a Cosine Similarity Loss using an axis of DEFAULT_AXIS, and a Loss Reduction of DEFAULT_REDUCTION
	CosineSimilarity(Ops tf, int axis) Creates a Cosine Similarity Loss using getSimpleName() as the loss name, and a Loss Reduction of DEFAULT_REDUCTION
	CosineSimilarity(Ops tf, int[] axis) Creates a Cosine Similarity Loss using getSimpleName() as the loss name, and a Loss Reduction of DEFAULT_REDUCTION
	CosineSimilarity(Ops tf, String name, int axis) Creates a Cosine Similarity Loss using a Loss Reduction of DEFAULT_REDUCTION
	CosineSimilarity(Ops tf, String name, int[] axis) Creates a Cosine Similarity Loss using a Loss Reduction of DEFAULT_REDUCTION
	CosineSimilarity(Ops tf, Reduction reduction) Creates a Cosine Similarity Loss using getSimpleName() as the loss name and an axis of DEFAULT_AXIS
	CosineSimilarity(Ops tf, String name, Reduction reduction) Creates a Cosine Similarity Loss using an axis of DEFAULT_AXIS
	CosineSimilarity(Ops tf, int axis, Reduction reduction) Creates a Cosine Similarity Loss using getSimpleName() as the loss name
	CosineSimilarity(Ops tf, int[] axis, Reduction reduction) Creates a Cosine Similarity Loss using getSimpleName() as the loss name
	CosineSimilarity(Ops tf, String name, int axis, Reduction reduction) Creates a Cosine Similarity Loss
	CosineSimilarity(Ops tf, String name, int[] axis, Reduction reduction) Creates a Cosine Similarity Loss

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

From class org.tensorflow.framework.losses.Loss

abstract <T extends TNumber> Operand<T>	call(Operand<? extends TNumber> labels, Operand<T> predictions, Operand<T> sampleWeights) Generates an Operand that calculates the loss.
<T extends TNumber> Operand<T>	call(Operand<? extends TNumber> labels, Operand<T> predictions) Calculates the loss
Reduction	getReduction() Gets the loss reduction
Ops	getTF() Gets the TensorFlow Ops

From class java.lang.Object

boolean	equals(Object arg0)
final Class<?>	getClass()
int	hashCode()
final void	notify()
final void	notifyAll()
String	toString()
final void	wait(long arg0, int arg1)
final void	wait(long arg0)
final void	wait()

Constants

public static final int DEFAULT_AXIS

Constant Value: -1

Fields

public static final Reduction DEFAULT_REDUCTION

Public Constructors

public CosineSimilarity (Ops tf)

Creates a Cosine Similarity Loss using getSimpleName() as the loss name, an axis of DEFAULT_AXIS, and a Loss Reduction of DEFAULT_REDUCTION

Parameters

tf	the TensorFlow Ops

public CosineSimilarity (Ops tf, String name)

Creates a Cosine Similarity Loss using an axis of DEFAULT_AXIS, and a Loss Reduction of DEFAULT_REDUCTION

Parameters

tf	the TensorFlow Ops
name	the name of the loss

public CosineSimilarity (Ops tf, int axis)

Creates a Cosine Similarity Loss using getSimpleName() as the loss name, and a Loss Reduction of DEFAULT_REDUCTION

Parameters

tf	the TensorFlow Ops
axis	The dimension along which the cosine similarity is computed.

public CosineSimilarity (Ops tf, int[] axis)

Creates a Cosine Similarity Loss using getSimpleName() as the loss name, and a Loss Reduction of DEFAULT_REDUCTION

Parameters

tf	the TensorFlow Ops
axis	The dimension along which the cosine similarity is computed.

public CosineSimilarity (Ops tf, String name, int axis)

Creates a Cosine Similarity Loss using a Loss Reduction of DEFAULT_REDUCTION

Parameters

tf	the TensorFlow Ops
name	the name of the loss
axis	The dimension along which the cosine similarity is computed.

public CosineSimilarity (Ops tf, String name, int[] axis)

Creates a Cosine Similarity Loss using a Loss Reduction of DEFAULT_REDUCTION

Parameters

tf	the TensorFlow Ops
name	the name of the loss
axis	The dimension along which the cosine similarity is computed.

public CosineSimilarity (Ops tf, Reduction reduction)

Creates a Cosine Similarity Loss using getSimpleName() as the loss name and an axis of DEFAULT_AXIS

Parameters

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

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

Creates a Cosine Similarity Loss using an axis of DEFAULT_AXIS

Parameters

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

public CosineSimilarity (Ops tf, int axis, Reduction reduction)

Creates a Cosine Similarity Loss using getSimpleName() as the loss name

Parameters

tf	the TensorFlow Ops
axis	The dimension along which the cosine similarity is computed.
reduction	Type of Reduction to apply to the loss.

public CosineSimilarity (Ops tf, int[] axis, Reduction reduction)

Creates a Cosine Similarity Loss using getSimpleName() as the loss name

Parameters

tf	the TensorFlow Ops
axis	The dimension along which the cosine similarity is computed.
reduction	Type of Reduction to apply to the loss.

public CosineSimilarity (Ops tf, String name, int axis, Reduction reduction)

Creates a Cosine Similarity Loss

Parameters

tf	the TensorFlow Ops
name	the name of the loss
axis	The dimension along which the cosine similarity is computed.
reduction	Type of Reduction to apply to the loss.

public CosineSimilarity (Ops tf, String name, int[] axis, Reduction reduction)

Creates a Cosine Similarity Loss

Parameters

tf	the TensorFlow Ops
name	the name of the loss
axis	The dimension along which the cosine similarity is computed.
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