Class Orthogonal<T extends TFloating>

java.lang.Object

org.tensorflow.framework.initializers.BaseInitializer<T>

org.tensorflow.framework.initializers.Orthogonal<T>

Type Parameters:

T - The TType for the call operation

All Implemented Interfaces:

Initializer<T>

public class Orthogonal<T extends TFloating>
extends BaseInitializer<T>

Initializer that generates an orthogonal matrix.

If the shape of the tensor to initialize is two-dimensional, it is initialized with an orthogonal matrix obtained from the QR decomposition of a matrix of random numbers drawn from a normal distribution. If the matrix has fewer rows than columns then the output will have orthogonal rows. Otherwise, the output will have orthogonal columns.

If the shape of the tensor to initialize is more than two-dimensional, a matrix of shape (shape.size(0) * ... * shape.size(n - 2), shape.size(n - 1)) is initialized, where n is the length of the shape vector. The matrix is subsequently reshaped to give a tensor of the desired shape.

Examples:

     Orthogonal<TFloat32, TFloat32> initializer =
             new org.tensorflow.framework.initializers.Orthogonal<>(tf);
     Operand<TFloat32> values =
             initializer.call(Ops tf, tf.constant(Shape.of(2,2)), TFloat32.class);

Field Summary

Fields

Modifier and Type	Field	Description
static final double	GAIN_DEFAULT

Constructor Summary

Constructors

Constructor	Description
Orthogonal(double gain, long seed)	Creates an Orthogonal Initializer
Orthogonal(long seed)	Creates an Orthogonal Initializer using GAIN_DEFAULT for the gain.

Method Summary

Modifier and Type	Method	Description
Operand<T>	call(Ops tf, Operand<TInt64> dims, Class<T> type)	Generates the operation used to perform the initialization.

Methods inherited from class BaseInitializer

getName

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

GAIN_DEFAULT

public static final double GAIN_DEFAULT

See Also:

• Constant Field Values

Constructor Details

Orthogonal

public Orthogonal(long seed)

Creates an Orthogonal Initializer using GAIN_DEFAULT for the gain.

Parameters:

seed - the seed for random number generation. An initializer created with a given seed will always produce the same random tensor for a given shape and dtype.

Orthogonal

public Orthogonal(double gain,
 long seed)

Creates an Orthogonal Initializer

Parameters:

gain - the gain to be applied to the Matrix.

seed - the seed for random number generation. An initializer created with a given seed will always produce the same random tensor for a given shape and dtype.

Method Details

call

public Operand<T> call(Ops tf,
 Operand<TInt64> dims,
 Class<T> type)

Generates the operation used to perform the initialization.

Parameters:

tf - the TensorFlow Ops

dims - the shape dimensions

type - the type of tensor

Returns:

An operand for the initialization.