JaccardLossLayer[]

represents a net layer that computes the Jaccard loss by comparing input class probability vectors with the target class vector.

JaccardLossLayer[n]

does the same but n defines the power of the denomenator, with n=2 the squared jaccard score is calculated.