Class VarianceScaling<T extends TFloating>

java.lang.Object

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

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

Type Parameters:

T - The TType for the call operation

All Implemented Interfaces:

Initializer<T>

Direct Known Subclasses:

Glorot, He, LeCun

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

Initializer capable of adapting its scale to the shape of weights tensors.

With distribution=TRUNCATED_NORMAL or NORMAL, samples are drawn from a truncated/untruncated normal distribution with a mean of zero and a standard deviation (after truncation, if used) stddev = Math.sqrt(scale / n), where n is:

• number of input units in the weight tensor, if mode=FAN_IN
• number of output units, if mode=FAN_OUT
• average of the numbers of input and output units, if mode=FAN_AVG

With distribution=UNIFORM, samples are drawn from a uniform distribution within [-limit, limit], where limit = Math.sqrt(3 * scale / n);.

Examples:

     long seed = 1234l;
     float scale = 0.1f;
     VarianceScaling<TFloat32, TFloat32> initializer =
         new org.tensorflow.framework.initializers.VarianceScaling<>(
             tf, scale, Mode.FAN_IN, Distribution.UNIFORM, seed);
     Operand<TFloat32> values =
         initializer.call(Ops tf, tf.constant(Shape.of(2,2)), TFloat32.class);

See Also:

• VarianceScaling.Mode
• VarianceScaling.Distribution

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static enum	VarianceScaling.Distribution	The random distribution to use when initializing the values.
static enum	VarianceScaling.Mode	The mode to use for calculating the fan values.

Field Summary

Fields

Modifier and Type	Field	Description
static final VarianceScaling.Distribution	DISTRIBUTION_DEFAULT
static final VarianceScaling.Mode	MODE_DEFAULT
static final double	SCALE_DEFAULT

Constructor Summary

Constructors

Constructor	Description
VarianceScaling(double scale, VarianceScaling.Mode mode, VarianceScaling.Distribution distribution, long seed)	Creates a VarianceScaling Initializer
VarianceScaling(long seed)	Creates a VarianceScaling Initializer

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

SCALE_DEFAULT

public static final double SCALE_DEFAULT

See Also:

• Constant Field Values

MODE_DEFAULT

public static final VarianceScaling.Mode MODE_DEFAULT

DISTRIBUTION_DEFAULT

public static final VarianceScaling.Distribution DISTRIBUTION_DEFAULT

Constructor Details

VarianceScaling

public VarianceScaling(long seed)

Creates a VarianceScaling Initializer

Parameters:

seed - sed to create random seeds.

VarianceScaling

public VarianceScaling(double scale,
 VarianceScaling.Mode mode,
 VarianceScaling.Distribution distribution,
 long seed)

Creates a VarianceScaling Initializer

Parameters:

scale - Scaling factor (positive float).

mode - the mode for the variance

distribution - Random distribution to use.

seed - Used to create random seeds.

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.