Accord.Math.Optimization.Losses Namespace 
Class  Description  

AbsoluteLoss 
Absolute loss, also known as L1loss.
 
AccuracyLoss 
Accuracy loss, also known as zerooneloss. This class
provides exactly the same functionality as ZeroOneLoss
but has a more intuitive name. Both classes are interchangeable.
 
BinaryCrossEntropyLoss 
Binary crossentropy loss for multilabel problems, also
known as logistic loss per output of a multilabel classifier.
 
CategoryCrossEntropyLoss 
Categorical crossentropy loss for multiclass problems,
also known as the logistic loss for softmax (categorical) outputs.
 
EuclideanLoss 
Euclidean loss, also known as zerooneloss. This class
provides exactly the same functionality as SquareLoss
but has a more intuitive name. Both classes are interchangeable.
 
HammingLoss 
Mean Accuracy loss, also known as zerooneloss per
class. Equivalent to ZeroOneLoss but
for multilabel classifiers.
 
LogLikelihoodLoss 
Negative loglikelihood loss.
 
LossBaseT 
Base class for loss functions.
 
LossBaseTInput, TScore, TLoss 
Base class for loss functions.
 
RSquaredLoss 
R² (rsquared) loss.
 
SquareLoss 
Square loss, also known as L2loss or Euclidean loss.
 
ZeroOneLoss 
Accuracy loss, also known as zerooneloss.

Structure  Description  

HingeLoss 
Hinge loss.
 
LogisticLoss 
Logistic loss.
 
SmoothHingeLoss 
Smooth Hinge loss.
 
SquaredHingeLoss 
Squared Hinge loss.

Interface  Description  

IDifferentiableLossTInput, TScore, TLoss 
Common interface for differentiable loss functions, such as SquareLoss,
HingeLoss, CategoryCrossEntropyLoss and
BinaryCrossEntropyLoss.
 
ILossT  
ILossTScore, TLoss 
Common interface for loss functions, such as SquareLoss,
HingeLoss, CategoryCrossEntropyLoss and
BinaryCrossEntropyLoss.
