wasserstein loss

loss = wasserstein_loss

Calculates the Wasserstein loss for a sample batch.

L1 loss

loss = l1loss

Calculate the L1 loss for a sample batch.

focal loss

loss = focal_loss:[gamma=2.0, alpha=0.25]

Calculate the focal loss for a sample batch, usually used to deal with class unbalanced problem.

margin loss

loss = margin_loss

Calculate the margin loss for a sample batch, a common used loss in Vector Capsule Networks with Dynanmic Routing.

spread loss

loss = "spreadloss"

Calculate the spread loss for a sample batch, a common used loss in Matrix Capsule Networks with EM Routing.