org.nd4j.autodiff.loss

Enum LossFunctions.Reduction

java.lang.Object

java.lang.Enum<LossFunctions.Reduction>

org.nd4j.autodiff.loss.LossFunctions.Reduction

All Implemented Interfaces:

Serializable, Comparable<LossFunctions.Reduction>

Enclosing class:

LossFunctions

public static enum LossFunctions.Reduction
extends Enum<LossFunctions.Reduction>

Enum Constant Summary

Enum Constants

Enum Constant and Description
MEAN_BY_COUNT Weighted mean: sum(weights * loss) / count(weights != 0) Example: 2d input, MSE loss along dimension 1.
MEAN_BY_WEIGHT Weighted mean: sum(weights * loss) / sum(weights) Example: 2d input, MSE loss along dimension 1.
NONE No reduction.
SPECIFIED_DIMS Reduce as normal along the specified dimensions, but don't sum/mean etc the remaining dimensions.
SUM Sum across the remaining dimensions, returning a scalar Example: 2d input, MSE loss along dimension 1.

Method Summary

Modifier and Type	Method and Description
static LossFunctions.Reduction	valueOf(String name) Returns the enum constant of this type with the specified name.
static LossFunctions.Reduction[]	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 LossFunctions.Reduction NONE

No reduction. Output is the same shape as the predictions/labels. Weights (if any) are applied. Dimension args are ignored. Example: 2d input, MSE. Output: sqDiff(predictions,labels) -> shape same as input/labels

SPECIFIED_DIMS

public static final LossFunctions.Reduction SPECIFIED_DIMS

Reduce as normal along the specified dimensions, but don't sum/mean etc the remaining dimensions. Example: 2d input, MSE loss along dimension 1. Output: mean(weights * sqDiff(predictions,labels),1) -> shape [dim0,1]

SUM

public static final LossFunctions.Reduction SUM

Sum across the remaining dimensions, returning a scalar Example: 2d input, MSE loss along dimension 1. Output: mse_per_ex = mean(weights * sqDiff(predictions,labels),1) *Same as SPECIFIED_DIMS* output = sum(mse_per_ex)

MEAN_BY_WEIGHT

public static final LossFunctions.Reduction MEAN_BY_WEIGHT

Weighted mean: sum(weights * loss) / sum(weights) Example: 2d input, MSE loss along dimension 1. Output: mse_per_ex = mean(weights * sqDiff(predictions,labels),1) *Same as SPECIFIED_DIMS* output = sum(mse_per_ex) / sum(weights) NOTE: if weights array is not provided, then weights default to (effectively) 1.0 for all entries - and hence MEAN_BY_WEIGHT is equivalent to SUM (as sum(1.0) = 1.0)

MEAN_BY_COUNT

public static final LossFunctions.Reduction MEAN_BY_COUNT

Weighted mean: sum(weights * loss) / count(weights != 0) Example: 2d input, MSE loss along dimension 1. Output: mse_per_ex = mean(weights * sqDiff(predictions,labels),1) *Same as SPECIFIED_DIMS* output = sum(mse_per_ex) / count(weights != 0) NOTE: if weights array is not provided, then weights default to scalar 1.0 and hence MEAN_BY_COUNT is equivalent to

Method Detail

values

public static LossFunctions.Reduction[] 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 (LossFunctions.Reduction c : LossFunctions.Reduction.values())
    System.out.println(c);

Returns:

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

valueOf

public static LossFunctions.Reduction valueOf(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:

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

NullPointerException - if the argument is null