Losses
These are all the available loss functions exported by ERM. See the corresponding usage page to understand how and when to use these functions.
AbsoluteLoss() constructs the l1/absolute loss. Use with Model().
HingeLoss() constructs the hinge loss (i.e., for SVM). Use with Model().
HuberLoss() constructs Huber loss. Use with Model(). Can also invoke as HuberLoss(alpha), which allows specification of the tradeoff parameter alpha > 0. Note that HuberLoss() defaults to alpha = 1.0.
LogisticLoss() constructs the logistic loss for classification problems. Use with Model()
SigmoidLoss() constructs the sigmoid loss for classification problems. Use with Model()
SquareLoss() constructs the l2/squared loss. Use with Model().