org.deeplearning4j.nn.conf

Enum GradientNormalization

java.lang.Object

java.lang.Enum<GradientNormalization>

org.deeplearning4j.nn.conf.GradientNormalization

All Implemented Interfaces:

java.io.Serializable, java.lang.Comparable<GradientNormalization>

public enum GradientNormalization
extends java.lang.Enum<GradientNormalization>

Gradient normalization strategies. These are applied on raw gradients, before the gradients are passed to the updater (SGD, RMSProp, Momentum, etc)

None = no gradient normalization (default)

RenormalizeL2PerLayer = rescale gradients by dividing by the L2 norm of all gradients for the layer.

RenormalizeL2PerParamType = rescale gradients by dividing by the L2 norm of the gradients, separately for each type of parameter within the layer.
This differs from RenormalizeL2PerLayer in that here, each parameter type (weight, bias etc) is normalized separately.
For example, in a MLP/FeedForward network (where G is the gradient vector), the output is as follows:

• GOut_weight = G_weight / l2(G_weight)
• GOut_bias = G_bias / l2(G_bias)

ClipElementWiseAbsoluteValue = clip the gradients on a per-element basis.
For each gradient g, set g <- sign(g)*max(maxAllowedValue,|g|).
i.e., if a parameter gradient has absolute value greater than the threshold, truncate it.
For example, if threshold = 5, then values in range -5<g<5 are unmodified; values <-5 are set to -5; values >5 are set to 5.
This was proposed by Mikolov (2012), Statistical Language Models Based on Neural Networks (thesis), http://www.fit.vutbr.cz/~imikolov/rnnlm/thesis.pdf in the context of learning recurrent neural networks.
Threshold for clipping can be set in Layer configuration, using gradientNormalizationThreshold(double threshold)

ClipL2PerLayer = conditional renormalization. Somewhat similar to RenormalizeL2PerLayer, this strategy scales the gradients if and only if the L2 norm of the gradients (for entire layer) exceeds a specified threshold. Specifically, if G is gradient vector for the layer, then:

• GOut = G if l2Norm(G) < threshold (i.e., no change)
• GOut = threshold * G / l2Norm(G) otherwise

Thus, the l2 norm of the scaled gradients will not exceed the specified threshold, though may be smaller than it
See: Pascanu, Mikolov, Bengio (2012), On the difficulty of training Recurrent Neural Networks, http://arxiv.org/abs/1211.5063
Threshold for clipping can be set in Layer configuration, using gradientNormalizationThreshold(double threshold)

ClipL2PerParamType = conditional renormalization. Very similar to ClipL2PerLayer, however instead of clipping per layer, do clipping on each parameter type separately.
For example in a recurrent neural network, input weight gradients, recurrent weight gradients and bias gradient are all clipped separately. Thus if one set of gradients are very large, these may be clipped while leaving the other gradients unmodified.
Threshold for clipping can be set in Layer configuration, using gradientNormalizationThreshold(double threshold)

Enum Constant Summary

Enum Constants

Enum Constant and Description
ClipElementWiseAbsoluteValue
ClipL2PerLayer
ClipL2PerParamType
None
RenormalizeL2PerLayer
RenormalizeL2PerParamType

Method Summary

Modifier and Type	Method and Description
static GradientNormalization	valueOf(java.lang.String name) Returns the enum constant of this type with the specified name.
static GradientNormalization[]	values() Returns an array containing the constants of this enum type, in the order they are declared.

Methods inherited from class java.lang.Enum

clone, compareTo, equals, finalize, getDeclaringClass, hashCode, name, ordinal, toString, valueOf

Methods inherited from class java.lang.Object

getClass, notify, notifyAll, wait, wait, wait

Enum Constant Detail

None

public static final GradientNormalization None

RenormalizeL2PerLayer

public static final GradientNormalization RenormalizeL2PerLayer

RenormalizeL2PerParamType

public static final GradientNormalization RenormalizeL2PerParamType

ClipElementWiseAbsoluteValue

public static final GradientNormalization ClipElementWiseAbsoluteValue

ClipL2PerLayer

public static final GradientNormalization ClipL2PerLayer

ClipL2PerParamType

public static final GradientNormalization ClipL2PerParamType

Method Detail

values

public static GradientNormalization[] values()

Returns an array containing the constants of this enum type, in the order they are declared. This method may be used to iterate over the constants as follows:

for (GradientNormalization c : GradientNormalization.values())
    System.out.println(c);

Returns:

an array containing the constants of this enum type, in the order they are declared

valueOf

public static GradientNormalization valueOf(java.lang.String name)

Returns the enum constant of this type with the specified name. The string must match exactly an identifier used to declare an enum constant in this type. (Extraneous whitespace characters are not permitted.)

Parameters:

name - the name of the enum constant to be returned.

Returns:

the enum constant with the specified name

Throws:

java.lang.IllegalArgumentException - if this enum type has no constant with the specified name

java.lang.NullPointerException - if the argument is null