Package ai.djl.training.loss

Class SoftmaxCrossEntropyLoss

java.lang.Object

ai.djl.training.evaluator.Evaluator

ai.djl.training.loss.Loss

ai.djl.training.loss.SoftmaxCrossEntropyLoss

public class SoftmaxCrossEntropyLoss
extends Loss

SoftmaxCrossEntropyLoss is a type of Loss that calculates the softmax cross entropy loss.

If sparse_label is true (default), label should contain integer category indicators. Then, \(L = -\sum_i \log p_{i, label_i}\). If sparse_label is false, label should be one-hot class coding or probability distribution and its shape should be the same as the shape of prediction. Then, \(L = -\sum_i \sum_j {label}_j \log p_{ij}\).

Field Summary

Fields inherited from class ai.djl.training.evaluator.Evaluator

totalInstances

Constructor Summary

Constructors

Constructor	Description
SoftmaxCrossEntropyLoss()	Creates a new instance of SoftmaxCrossEntropyLoss with default parameters.
SoftmaxCrossEntropyLoss(java.lang.String name)	Creates a new instance of SoftmaxCrossEntropyLoss with default parameters.
SoftmaxCrossEntropyLoss(java.lang.String name, float weight, int classAxis, boolean sparseLabel, boolean fromLogit)	Creates a new instance of SoftmaxCrossEntropyLoss with the given parameters.

Method Summary

Modifier and Type	Method	Description
NDArray	evaluate(NDList label, NDList prediction)	Calculates the evaluation between the labels and the predictions.

Methods inherited from class ai.djl.training.loss.Loss

addAccumulator, elasticNetWeightedDecay, elasticNetWeightedDecay, elasticNetWeightedDecay, elasticNetWeightedDecay, getAccumulator, hingeLoss, hingeLoss, hingeLoss, l1Loss, l1Loss, l1Loss, l1WeightedDecay, l1WeightedDecay, l1WeightedDecay, l2Loss, l2Loss, l2Loss, l2WeightedDecay, l2WeightedDecay, l2WeightedDecay, maskedSoftmaxCrossEntropyLoss, maskedSoftmaxCrossEntropyLoss, maskedSoftmaxCrossEntropyLoss, quantileL1Loss, quantileL1Loss, resetAccumulator, sigmoidBinaryCrossEntropyLoss, sigmoidBinaryCrossEntropyLoss, sigmoidBinaryCrossEntropyLoss, softmaxCrossEntropyLoss, softmaxCrossEntropyLoss, softmaxCrossEntropyLoss, updateAccumulator

Methods inherited from class ai.djl.training.evaluator.Evaluator

checkLabelShapes, checkLabelShapes, getName

Methods inherited from class java.lang.Object

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

Constructor Detail

SoftmaxCrossEntropyLoss

public SoftmaxCrossEntropyLoss()

Creates a new instance of SoftmaxCrossEntropyLoss with default parameters.

SoftmaxCrossEntropyLoss

public SoftmaxCrossEntropyLoss(java.lang.String name)

Creates a new instance of SoftmaxCrossEntropyLoss with default parameters.

Parameters:

name - the name of the loss

SoftmaxCrossEntropyLoss

public SoftmaxCrossEntropyLoss(java.lang.String name,
                               float weight,
                               int classAxis,
                               boolean sparseLabel,
                               boolean fromLogit)

Creates a new instance of SoftmaxCrossEntropyLoss with the given parameters.

Parameters:

name - the name of the loss

weight - the weight to apply on the loss value, default 1

classAxis - the axis that represents the class probabilities, default -1

sparseLabel - whether labels are rank-1 integer array of [batch_size] (true) or rank-2 one-hot or probability distribution of shape [batch_size, n-class] (false), default true

fromLogit - if true, the inputs are assumed to be the numbers before being applied with softmax. Then logSoftmax will be applied to input, default true

Method Detail

evaluate

public NDArray evaluate(NDList label,
                        NDList prediction)

Calculates the evaluation between the labels and the predictions.

Specified by:

evaluate in class Evaluator

Parameters:

label - the correct values

prediction - the predicted values

Returns:

the evaluation result